Search

GPT-4o is an autoregressive omni model that accepts as input any combination of text, audio, image, and video, and generates any combination of text, audio, and image outputs. It's trained end-to-end across text, vision, and audio, meaning all inputs and outputs are processed by the same neural network. GPT-4o can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time in conversation. It matches GPT-4 Turbo performance on text in English and code, with significant improvement on text in non-English languages, while also being much faster and 50\% cheaper in the API. GPT-4o is especially better at vision and audio understanding compared to existing models. In line with our commitment to building AI safely and consistent with our voluntary commitments to the White House, we are sharing the GPT-4o System Card, which includes our Preparedness Framework evaluations. In this System Card, we provide a detailed look at GPT-4o's capabilities, limitations, and safety evaluations across multiple categories, focusing on speech-to-speech while also evaluating text and image capabilities, and measures we've implemented to ensure the model is safe and aligned. We also include third-party assessments on dangerous capabilities, as well as discussion of potential societal impacts of GPT-4o's text and vision capabilities.

In this work, we present MedImageInsight, an open-source medical imaging embedding model. MedImageInsight is trained on medical images with associated text and labels across a diverse collection of domains, including X-Ray, CT, MRI, dermoscopy, OCT, fundus photography, ultrasound, histopathology, and mammography. Rigorous evaluations demonstrate MedImageInsight's ability to achieve state-of-the-art (SOTA) or human expert level performance across classification, image-image search, and fine-tuning tasks. Specifically, on public datasets, MedImageInsight achieves SOTA in CT 3D medical image retrieval, as well as SOTA in disease classification and search for chest X-ray, dermatology, and OCT imaging. Furthermore, MedImageInsight achieves human expert performance in bone age estimation (on both public and partner data), as well as AUC above 0.9 in most other domains. When paired with a text decoder, MedImageInsight achieves near SOTA level single image report findings generation with less than 10\% the parameters of other models. Compared to fine-tuning GPT-4o with only MIMIC-CXR data for the same task, MedImageInsight outperforms in clinical metrics, but underperforms on lexical metrics where GPT-4o sets a new SOTA. Importantly for regulatory purposes, MedImageInsight can generate ROC curves, adjust sensitivity and specificity based on clinical need, and provide evidence-based decision support through image-image search (which can also enable retrieval augmented generation). In an independent clinical evaluation of image-image search in chest X-ray, MedImageInsight outperformed every other publicly available foundation model evaluated by large margins (over 6 points AUC), and significantly outperformed other models in terms of AI fairness (across age and gender). We hope releasing MedImageInsight will help enhance collective progress in medical imaging AI research and development.

Large Language Models (LLMs) have excelled in multi-hop question-answering (M-QA) due to their advanced reasoning abilities. However, the impact of the inherent reasoning structures on LLM M-QA performance remains unclear, largely due to the absence of QA datasets that provide fine-grained reasoning structures. To address this gap, we introduce the Graph Reasoning-Structured Question Answering Dataset (GRS-QA), which includes both semantic contexts and reasoning structures for QA pairs. Unlike existing M-QA datasets, where different reasoning structures are entangled together, GRS-QA explicitly captures intricate reasoning pathways by constructing reasoning graphs, where nodes represent textual contexts and edges denote logical flows. These reasoning graphs of different structures enable a fine-grained evaluation of LLM reasoning capabilities across various reasoning structures. Our empirical analysis reveals that LLMs perform differently when handling questions with varying reasoning structures. This finding facilitates the exploration of textual structures as compared with semantics., Comment: 15 pages, 24 figures, 10 tables

Instruction tuning is a supervised fine-tuning approach that significantly improves the ability of large language models (LLMs) to follow human instructions. We propose SelfCodeAlign, the first fully transparent and permissive pipeline for self-aligning code LLMs without extensive human annotations or distillation. SelfCodeAlign employs the same base model for inference throughout the data generation process. It first extracts diverse coding concepts from high-quality seed snippets to generate new tasks. It then samples multiple responses per task, pairs each with test cases, and validates them in a sandbox environment. Finally, passing examples are selected for instruction tuning. In our primary experiments, we use SelfCodeAlign with CodeQwen1.5-7B to generate a dataset of 74k instruction-response pairs. Finetuning on this dataset leads to a model that achieves a 67.1 pass@1 on HumanEval+, surpassing CodeLlama-70B-Instruct despite being ten times smaller. Across all benchmarks, this finetuned model consistently outperforms the original version trained with OctoPack, the previous state-of-the-art method for instruction tuning without human annotations or distillation. Additionally, we show that SelfCodeAlign is effective across LLMs of various sizes, from 3B to 33B, and that the base models can benefit more from alignment with their own data distribution. We further validate each component's effectiveness in our pipeline, showing that SelfCodeAlign outperforms both direct distillation from GPT-4o and leading GPT-3.5-based distillation methods, such as OSS-Instruct and Evol-Instruct. SelfCodeAlign has also led to the creation of StarCoder2-Instruct, the first fully transparent, permissively licensed, and self-aligned code LLM that achieves state-of-the-art coding performance., Comment: Accepted to NeurIPS 2024

Eye tracking technology offers great potential for improving road safety. It is already being built into vehicles, namely cars and trucks. When this technology is integrated into transit service vehicles, employees, i.e., bus drivers, will be subject to being eye tracked on their job. Although there is much research effort advancing algorithms for eye tracking in transportation, less is known about how end users perceive this technology, especially when interacting with it in an employer-mandated context. In this first study of its kind, we investigated transit bus operators' perceptions of eye tracking technology. From a methodological perspective, we introduce a mixed methods approach where participants experience the technology first-hand and then reflect on their experience while viewing a playback of the recorded data. Thematic analysis of the interview transcripts reveals interesting potential uses of eye tracking in this work context and surfaces transit operators' fears and concerns about this technology.

Consider shuffling a deck of $n$ cards, labeled $1$ through $n$, as follows: at each time step, pick one card uniformly with your right hand and another card, independently and uniformly with your left hand; then swap the cards. How long does it take until the deck is close to random? Diaconis and Shahshahani showed that this process undergoes cutoff in total variation distance at time $t = \lfloor n\log{n}/2 \rfloor$. Confirming a conjecture of N.~Berestycki, we prove the definitive ``hitting time'' version of this result: let $\tau$ denote the first time at which all cards have been touched. The total variation distance between the stopped distribution at $\tau$ and the uniform distribution on permutations is $o_n(1)$; this is best possible, since at time $\tau-1$, the total variation distance is at least $(1+o_n(1))e^{-1}$., Comment: 15 pages

Solar magnetic activity follows regular cycles of about 11 years with an inversion of polarity in the poles every 22 years. This changing surface magnetism impacts the properties of the acoustic modes. The acoustic mode frequency shifts are a good proxy of the magnetic cycle. In this Letter we investigate solar magnetic activity cycles 23 and 24 through the evolution of the frequency shifts of low-degree modes (l= 0, 1, and 2) in three frequency bands. These bands probe properties between 74 and 1575 km beneath the surface. The analysis was carried out using observations from the space instrument Global Oscillations at Low Frequency and the ground-based Birmingham Solar Oscillations Network and Global Oscillation Network Group. The frequency shifts of radial modes suggest that changes in the magnetic field amplitude and configuration likely occur near the Sun's surface rather than near its core. The maximum shifts of solar cycle 24 occurred earlier at mid and high latitudes (relative to the equator) and about 1550 km beneath the photosphere. At this depth but near the equator, this maximum aligns with the surface activity but has a stronger magnitude. At around 74 km deep, the behaviour near the equator mirrors the behaviour at the surface, while at higher latitudes, it matches the strength of cycle 23., Comment: Letter accepted in Astronomy and Astrophysics, 10 pages, 5 figures

Agent learning from human interaction often relies on explicit signals, but implicit social cues, such as prosody in speech, could provide valuable information for more effective learning. This paper advocates for the integration of prosody as a teaching signal to enhance agent learning from human teachers. Through two exploratory studies--one examining voice feedback in an interactive reinforcement learning setup and the other analyzing restricted audio from human demonstrations in three Atari games--we demonstrate that prosody carries significant information about task dynamics. Our findings suggest that prosodic features, when coupled with explicit feedback, can enhance reinforcement learning outcomes. Moreover, we propose guidelines for prosody-sensitive algorithm design and discuss insights into teaching behavior. Our work underscores the potential of leveraging prosody as an implicit signal for more efficient agent learning, thus advancing human-agent interaction paradigms., Comment: Published at the 26th ACM International Conference on Multimodal Interaction (ICMI) 2024

We report the discovery of the repeating fast radio burst source FRB 20240209A using the CHIME/FRB telescope. We have detected 22 bursts from this repeater between February and July 2024, six of which were also recorded at the Outrigger station KKO. The 66-km long CHIME-KKO baseline can provide single-pulse FRB localizations along one dimension with $2^{\prime\prime}$ accuracy. The high declination of $\sim$86 degrees for this repeater allowed its detection with a rotating range of baseline vectors, enabling the combined localization region size to be constrained to $1^{\prime\prime}\times2^{\prime\prime}$. We present deep Gemini observations that, combined with the FRB localization, enabled a robust association of FRB 20240209A to the outskirts of a luminous galaxy (P(O|x) = 0.99; $L \approx 5.3 \times 10^{10}\,L_{\odot}$). FRB 20240209A has a projected physical offset of $40 \pm 5$ kpc from the center of its host galaxy, making it the FRB with the largest host galaxy offset to date. When normalized by the host galaxy size, the offset of FRB 20240209A is comparable to that of FRB 20200120E, the only FRB source known to originate in a globular cluster. We consider several explanations for the large offset, including a progenitor that was kicked from the host galaxy or in situ formation in a low-luminosity satellite galaxy of the putative host, but find the most plausible scenario to be a globular cluster origin. This, coupled with the quiescent, elliptical nature of the host as demonstrated in our companion paper, provide strong evidence for a delayed formation channel for the progenitor of the FRB source., Comment: Submitted to AAS Journals

Respiratory illnesses are a significant global health burden. Respiratory illnesses, primarily Chronic obstructive pulmonary disease (COPD), is the seventh leading cause of poor health worldwide and the third leading cause of death worldwide, causing 3.23 million deaths in 2019, necessitating early identification and diagnosis for effective mitigation. Among the diagnostic tools employed, spirometry plays a crucial role in detecting respiratory abnormalities. However, conventional clinical spirometry methods often entail considerable costs and practical limitations like the need for specialized equipment, trained personnel, and a dedicated clinical setting, making them less accessible. To address these challenges, wearable spirometry technologies have emerged as promising alternatives, offering accurate, cost-effective, and convenient solutions. The development of machine learning models for wearable spirometry heavily relies on the availability of high-quality ground truth spirometry data, which is a laborious and expensive endeavor. In this research, we propose using active learning, a sub-field of machine learning, to mitigate the challenges associated with data collection and labeling. By strategically selecting samples from the ground truth spirometer, we can mitigate the need for resource-intensive data collection. We present evidence that models trained on small subsets obtained through active learning achieve comparable/better results than models trained on the complete dataset.

In the last years there has been a growing interest in the study of learning problems associated with algebraic structures. The framework we use models the scenario in which a learner is given larger and larger fragments of a structure from a given target family and is required to output an hypothesis about the structure's isomorphism type. So far researchers focused on $\mathbf{Ex}$-learning, in which the learner is asked to eventually stabilize to the correct hypothesis, and on restrictions where the learner is allowed to change the hypothesis a fixed number of times. Yet, other learning paradigms coming from classical algorithmic learning theory remained unexplored. We study the "learning power" of such criteria, comparing them via descriptive-set-theoretic tools thanks to the novel notion of $E$-learnability. The main outcome of this paper is that such criteria admit natural syntactic characterizations in terms of infinitary formulas analogous to the one given for $\mathbf{Ex}$-learning in [6]. Such characterizations give a powerful method to understand whether a family of structure is learnable with respect to the desired criterion.

Single crystalline materials, different from polycrystalline and twinning structures, are desired for investigating the intrinsic physical properties, as grain and twin boundaries often work as a source of artifacts. Bismuth chalcogenides, which are van der Waals materials notable as topological insulators, have attracted significant interest due to their rich physical properties. However, the formation of 60{\deg} twin domains is common in these materials. Here, we demonstrate single-domain bismuth chalcogenides. Using atomic force microscopy, we investigated the morphology of Bi2Se3 and Bi2Te3 grown on Al2O3. Despite lattice constants of Bi2Se3 and Al2O3 substrates being well matched with hybrid symmetry epitaxy, Bi2Se3 exhibited 60{\deg} twin boundaries across the surface. Interestingly, Bi2Te3 showed a single-domain feature across the 10 mm by 10 mm sample even with lattice mismatch. While further in-depth studies are required to understand this difference in the morphology between Bi2Se3/Al2O3 and Bi2Te3/Al2O3, we suggest that the formation of twin boundaries in bismuth chalcogenides is related to the interaction between quintuple layers across the van der Waals gap rather than strain or defects.

Personalized recommendation is a ubiquitous application on the internet, with many industries and hyperscalers extensively leveraging Deep Learning Recommendation Models (DLRMs) for their personalization needs (like ad serving or movie suggestions). With growing model and dataset sizes pushing computation and memory requirements, GPUs are being increasingly preferred for executing DLRM inference. However, serving newer DLRMs, while meeting acceptable latencies, continues to remain challenging, making traditional deployments increasingly more GPU-hungry, resulting in higher inference serving costs. In this paper, we show that the embedding stage continues to be the primary bottleneck in the GPU inference pipeline, leading up to a 3.2x embedding-only performance slowdown. To thoroughly grasp the problem, we conduct a detailed microarchitecture characterization and highlight the presence of low occupancy in the standard embedding kernels. By leveraging direct compiler optimizations, we achieve optimal occupancy, pushing the performance by up to 53%. Yet, long memory latency stalls continue to exist. To tackle this challenge, we propose specialized plug-and-play-based software prefetching and L2 pinning techniques, which help in hiding and decreasing the latencies. Further, we propose combining them, as they complement each other. Experimental evaluations using A100 GPUs with large models and datasets show that our proposed techniques improve performance by up to 103% for the embedding stage, and up to 77% for the overall DLRM inference pipeline., Comment: This work has been accepted in the 57th MICRO (https://microarch.org/micro57/program/). Please check appendix for details on reproducing our work including codebase and steps

A variety of neural networks architectures are being studied to tackle blur in images and videos caused by a non-steady camera and objects being captured. In this paper, we present an overview of these existing networks and perform experiments to remove the blur caused by atmospheric turbulence. Our experiments aim to examine the reusability of existing networks and identify desirable aspects of the architecture in a system that is geared specifically towards atmospheric turbulence mitigation. We compare five different architectures, including a network trained in an end-to-end fashion, thereby removing the need for a stabilization step.

Mobile health (mHealth) programs face a critical challenge in optimizing the timing of automated health information calls to beneficiaries. This challenge has been formulated as a collaborative multi-armed bandit problem, requiring online learning of a low-rank reward matrix. Existing solutions often rely on heuristic combinations of offline matrix completion and exploration strategies. In this work, we propose a principled Bayesian approach using Thompson Sampling for this collaborative bandit problem. Our method leverages prior information through efficient Gibbs sampling for posterior inference over the low-rank matrix factors, enabling faster convergence. We demonstrate significant improvements over state-of-the-art baselines on a real-world dataset from the world's largest maternal mHealth program. Our approach achieves a $16\%$ reduction in the number of calls compared to existing methods and a $47$\% reduction compared to the deployed random policy. This efficiency gain translates to a potential increase in program capacity by $0.5-1.4$ million beneficiaries, granting them access to vital ante-natal and post-natal care information. Furthermore, we observe a $7\%$ and $29\%$ improvement in beneficiary retention (an extremely hard metric to impact) compared to state-of-the-art and deployed baselines, respectively. Synthetic simulations further demonstrate the superiority of our approach, particularly in low-data regimes and in effectively utilizing prior information. We also provide a theoretical analysis of our algorithm in a special setting using Eluder dimension.

The introduction of a twist between two layers of two-dimensional materials has opened up a new and exciting field of research known as twistronics. In these systems, the phonon dispersions show significant renormalization and enhanced electron-phonon interactions as a function of the twist angle. However, the large system size of the resulting moir\'e patterns in these systems makes phonon calculations computationally challenging. In this paper, we present PARPHOM, a powerful code package designed to address these challenges. PARPHOM enables the generation of force constants, computation of phononic band structures, and determination of density of states in twisted 2D material systems. Moreover, PARPHOM provides essential routines to investigate the finite temperature dynamics in these systems and analyze the chirality of the phonon bands. This paper serves as an introduction to PARPHOM, highlighting its capabilities and demonstrating its utility in unraveling the intricate phononic properties of twisted 2D materials., Comment: 10 pages, 7 figures

This study examines the relationship between early solar coronal mass ejection (CME) propagation, the associated filament eruption, and coronal dimming in the rare event observed on March 28, 2022, which featured a three-part CME in the low corona of active region AR 12975, including a bright core/filament, dark cavity, and bright front edge. We employ 3D filament and CME shock reconstructions using data from SolO, STEREO-A, and SDO to track the filament's path, height, and kinematics. Our analysis across three viewpoints shows the outer front in SolO/EUI 304 \r{A} aligns with shock structures in STEREO-A/EUVI 195 \r{A}, showing a full 3D EUV wave dome, later matching the outer CME front in STEREO-A COR2. We introduce the method ATLAS-3D (Advanced Technique for single Line-of-sight Acquisition of Structures in 3D) and validate it against traditional approaches to reconstruct CME shock using SOLO data exclusively. Additionally, we estimate early CME propagation characteristics based on coronal dimming evolution with the DIRECD method. Results indicate that the filament height increased from 28 to 616 Mm (0.04 to 0.89 Rs) within 30 minutes (11:05 to 11:35 UT), reaching peak velocity of around 648 km/s and acceleration of around 1624 m/s$^2$. At 11:45 UT, the filament deflected by 12{\deg} to a height of 841 Mm (1.21 Rs), while the CME shock expanded from 383 to 837 Mm (0.55 to 1.2 Rs) over 10 minutes. Key parameters include a CME direction inclined by 6{\deg}, a 21{\deg} half-width, and a 1.12 Rs cone height at the dimming's impulsive phase end. This event demonstrates that expanding dimming correlates with early CME development, with the DIRECD method linking 2D dimming to 3D CME evolution. These insights underscore the value of multi-viewpoint observations and advanced reconstructions for improving space weather forecasting., Comment: 14 pages, 9 main figures, 2 appendix figures, 1 table, accepted for Publication in Astronomy and Astrophysics

With the rapid growth of social media usage, a common trend has emerged where users often make sarcastic comments on posts. While sarcasm can sometimes be harmless, it can blur the line with cyberbullying, especially when used in negative or harmful contexts. This growing issue has been exacerbated by the anonymity and vast reach of the internet, making cyberbullying a significant concern on platforms like Reddit. Our research focuses on distinguishing cyberbullying from sarcasm, particularly where online language nuances make it difficult to discern harmful intent. This study proposes a framework using natural language processing (NLP) and machine learning to differentiate between the two, addressing the limitations of traditional sentiment analysis in detecting nuanced behaviors. By analyzing a custom dataset scraped from Reddit, we achieved a 95.15% accuracy in distinguishing harmful content from sarcasm. Our findings also reveal that teenagers and minority groups are particularly vulnerable to cyberbullying. Additionally, our research uncovers coordinated graphs of groups involved in cyberbullying, identifying common patterns in their behavior. This research contributes to improving detection capabilities for safer online communities., Comment: 7 pages, 4 figures

This paper considers the problem of annotating datapoints using an expert with only a few annotation rounds in a label-scarce setting. We propose soliciting reliable feedback on difficulty in annotating a datapoint from the expert in addition to ground truth label. Existing literature in active learning or coreset selection turns out to be less relevant to our setting since they presume the existence of a reliable trained model, which is absent in the label-scarce regime. However, the literature on coreset selection emphasizes the presence of difficult data points in the training set to perform supervised learning in downstream tasks (Mindermann et al., 2022). Therefore, for a given fixed annotation budget of $\mathsf{T}$ rounds, we model the sequential decision-making problem of which (difficult) datapoints to choose for annotation in a sparse linear bandits framework with the constraint that no arm can be pulled more than once (blocking constraint). With mild assumptions on the datapoints, our (computationally efficient) Explore-Then-Commit algorithm BSLB achieves a regret guarantee of $\widetilde{\mathsf{O}}(k^{\frac{1}{3}} \mathsf{T}^{\frac{2}{3}} +k^{-\frac{1}{2}} \beta_k + k^{-\frac{1}{12}} \beta_k^{\frac{1}{2}}\mathsf{T}^{\frac{5}{6}})$ where the unknown parameter vector has tail magnitude $\beta_k$ at sparsity level $k$. To this end, we show offline statistical guarantees of Lasso estimator with mild Restricted Eigenvalue (RE) condition that is also robust to sparsity. Finally, we propose a meta-algorithm C-BSLB that does not need knowledge of the optimal sparsity parameters at a no-regret cost. We demonstrate the efficacy of our BSLB algorithm for annotation in the label-scarce setting for an image classification task on the PASCAL-VOC dataset, where we use real-world annotation difficulty scores., Comment: 31 Pages

Large Language Models (LLMs) and Text-To-Image (T2I) models have demonstrated the ability to generate compelling text and visual stories. However, their outputs are predominantly aligned with the sensibilities of the Global North, often resulting in an outsider's gaze on other cultures. As a result, non-Western communities have to put extra effort into generating culturally specific stories. To address this challenge, we developed a visual storytelling pipeline called KAHANI that generates culturally grounded visual stories for non-Western cultures. Our pipeline leverages off-the-shelf models GPT-4 Turbo and Stable Diffusion XL (SDXL). By using Chain of Thought (CoT) and T2I prompting techniques, we capture the cultural context from user's prompt and generate vivid descriptions of the characters and scene compositions. To evaluate the effectiveness of KAHANI, we conducted a comparative user study with ChatGPT-4 (with DALL-E3) in which participants from different regions of India compared the cultural relevance of stories generated by the two tools. Results from the qualitative and quantitative analysis performed on the user study showed that KAHANI was able to capture and incorporate more Culturally Specific Items (CSIs) compared to ChatGPT-4. In terms of both its cultural competence and visual story generation quality, our pipeline outperformed ChatGPT-4 in 27 out of the 36 comparisons., Comment: Under review

On May 8, 2024, solar active region 13664 produced an X-class flare, several M-class flares, and multiple Earth-directed Coronal Mass Ejections (CMEs). The initial CME caused coronal dimmings, characterized by localized reductions in extreme-ultraviolet (EUV) emissions, indicating mass loss and expansion during the eruption. After one solar rotation, on June 8, 2024, the same region produced another M-class flare followed by coronal dimmings observed by the SDO and STEREO spacecraft. We analyzed early CME evolution and direction from coronal dimming expansion at the end of the impulsive phase using the DIRECD (Dimming Inferred Estimation of CME Direction) method. To validate the 3D CME cone, we compared CME properties from the low corona with white-light coronagraph data. The May 8 CME expanded radially, with a 7.7 deg inclination, 70 deg angular width, and 0.81 Rsun cone height, while the June 8 CME had a 15.7 deg inclination, 81 deg width, and 0.89 Rsun height. Our study shows that tracking low coronal signatures, like coronal dimming expansion, can estimate CME direction early, providing crucial lead time for space weather forecasts., Comment: Accepted for Publication in Astronomy and Astrophysics. arXiv admin note: text overlap with arXiv:2311.13942

Dynamic Nuclear Polarization (DNP) is transforming NMR and MRI by significantly enhancing sensitivity through the transfer of polarization from electron spins to nuclear spins via microwave irradiation. However, the use of monochromatic continuous-wave (CW) irradiation limits the efficiency of DNP for systems with heterogeneous broad EPR lines. Broad-band techniques such as chirp irradiation offer a solution, particularly for Solid Effect (SE) DNP in such cases. Despite its widespread use, the role of quantum coherence generated during chirp irradiation remains unclear, even though it is a key factor in determining the maximum achievable DNP efficiency. In this work, we use density matrix formalism to provide a comprehensive understanding of the quantum coherence generated during electron-nucleus double-quantum (DQ) and zero-quantum (ZQ) SE transitions and their impact on Integrated Solid Effect (ISE) DNP under chirp irradiation. Our analysis employs fictitious product-operator bases to trace the evolution of electron-nucleus coherence leading to integrated or differentiated SE. We also explore the previously unexamined role of decoherence in optimizing chirped DNP. These findings provide new insights into low-temperature DNP and triplet DNP using photoexcited electrons.

We propose Audio Noise Awareness using Visuals of Indoors for NAVIgation for quieter robot path planning. While humans are naturally aware of the noise they make and its impact on those around them, robots currently lack this awareness. A key challenge in achieving audio awareness for robots is estimating how loud will the robot's actions be at a listener's location? Since sound depends upon the geometry and material composition of rooms, we train the robot to passively perceive loudness using visual observations of indoor environments. To this end, we generate data on how loud an 'impulse' sounds at different listener locations in simulated homes, and train our Acoustic Noise Predictor (ANP). Next, we collect acoustic profiles corresponding to different actions for navigation. Unifying ANP with action acoustics, we demonstrate experiments with wheeled (Hello Robot Stretch) and legged (Unitree Go2) robots so that these robots adhere to the noise constraints of the environment. See code and data at https://anavi-corl24.github.io/, Comment: 8th Conference on Robot Learning (CoRL) 2024

Since the discovery of superconductivity in the Fe(Te,Se) system, it has been a general consensus that the end member of FeTe is not superconducting. Nonetheless, in recent years, there have been reports of superconducting FeTe films, but the origin of their superconductivity remains mysterious. Here, we provide the first comprehensive review of all the reported FeTe films regarding the relationship between their superconductivity and neighboring layers. Based on this review, we show that telluride neighboring layers are the key to superconducting FeTe films. Then, with additional new studies, we show that stoichiometric Te content, which can be readily achieved in FeTe films with the assistance of neighboring telluride layers, might be crucial to stabilizing the superconductivity in this system. This work provides insights into the underlying mechanism behind superconductivity in FeTe films and sheds light on the critical role of neighboring layers and stoichiometry control toward manipulating topological superconductivity in FeTe heterostructures., Comment: 16 pages, 4 figures

Graph distillation has emerged as a promising avenue to enable scalable training of GNNs by compressing the training dataset while preserving essential graph characteristics. Our study uncovers significant shortcomings in current graph distillation techniques. First, the majority of the algorithms paradoxically require training on the full dataset to perform distillation. Second, due to their gradient-emulating approach, these methods require fresh distillation for any change in hyperparameters or GNN architecture, limiting their flexibility and reusability. Finally, they fail to achieve substantial size reduction due to synthesizing fully-connected, edge-weighted graphs. To address these challenges, we present Bonsai, a novel graph distillation method empowered by the observation that \textit{computation trees} form the fundamental processing units of message-passing GNNs. Bonsai distills datasets by encoding a careful selection of \textit{exemplar} trees that maximize the representation of all computation trees in the training set. This unique approach imparts Bonsai as the first linear-time, model-agnostic graph distillation algorithm for node classification that outperforms existing baselines across $6$ real-world datasets on accuracy, while being $22$ times faster on average. Bonsai is grounded in rigorous mathematical guarantees on the adopted approximation strategies making it robust to GNN architectures, datasets, and parameters.

Machine-generated data is a valuable resource for training Artificial Intelligence algorithms, evaluating rare workflows, and sharing data under stricter data legislations. The challenge is to generate data that is accurate and private. Current statistical and deep learning methods struggle with large data volumes, are prone to hallucinating scenarios incompatible with reality, and seldom quantify privacy meaningfully. Here we introduce Genomator, a logic solving approach (SAT solving), which efficiently produces private and realistic representations of the original data. We demonstrate the method on genomic data, which arguably is the most complex and private information. Synthetic genomes hold great potential for balancing underrepresented populations in medical research and advancing global data exchange. We benchmark Genomator against state-of-the-art methodologies (Markov generation, Restricted Boltzmann Machine, Generative Adversarial Network and Conditional Restricted Boltzmann Machines), demonstrating an 84-93% accuracy improvement and 95-98% higher privacy. Genomator is also 1000-1600 times more efficient, making it the only tested method that scales to whole genomes. We show the universal trade-off between privacy and accuracy, and use Genomator's tuning capability to cater to all applications along the spectrum, from provable private representations of sensitive cohorts, to datasets with indistinguishable pharmacogenomic profiles. Demonstrating the production-scale generation of tuneable synthetic data can increase trust and pave the way into the clinic.

Enabling clinicians and researchers to directly interact with global genomic data resources by removing technological barriers is vital for medical genomics. AskBeacon enables Large Language Models to be applied to securely shared cohorts via the GA4GH Beacon protocol. By simply "asking" Beacon, actionable insights can be gained, analyzed and made publication-ready.

We consider Directed Steiner Forest (DSF), a fundamental problem in network design. The input to DSF is a directed edge-weighted graph $G = (V, E)$ and a collection of vertex pairs $\{(s_i, t_i)\}_{i \in [k]}$. The goal is to find a minimum cost subgraph $H$ of $G$ such that $H$ contains an $s_i$-$t_i$ path for each $i \in [k]$. DSF is NP-Hard and is known to be hard to approximate to a factor of $\Omega(2^{\log^{1 - \epsilon}(n)})$ for any fixed $\epsilon > 0$ [DK'99]. DSF admits approximation ratios of $O(k^{1/2 + \epsilon})$ [CEGS'11] and $O(n^{2/3 + \epsilon})$ [BBMRY'13]. In this work we show that in planar digraphs, an important and useful class of graphs in both theory and practice, DSF is much more tractable. We obtain an $O(\log^6 k)$-approximation algorithm via the junction tree technique. Our main technical contribution is to prove the existence of a low density junction tree in planar digraphs. To find an approximate junction tree we rely on recent results on rooted directed network design problems [FM'23, CJKZZ'24], in particular, on an LP-based algorithm for the Directed Steiner Tree problem [CJKZZ'24]. Our work and several other recent ones on algorithms for planar digraphs [FM'23, KS'21, CJKZZ'24] are built upon structural insights on planar graph reachability and shortest path separators [Thorup'04].

We present the results of a search for gravitational-wave transients associated with core-collapse supernova SN 2023ixf, which was observed in the galaxy Messier 101 via optical emission on 2023 May 19th, during the LIGO-Virgo-KAGRA 15th Engineering Run. We define a five-day on-source window during which an accompanying gravitational-wave signal may have occurred. No gravitational waves have been identified in data when at least two gravitational-wave observatories were operating, which covered $\sim 14\%$ of this five-day window. We report the search detection efficiency for various possible gravitational-wave emission models. Considering the distance to M101 (6.7 Mpc), we derive constraints on the gravitational-wave emission mechanism of core-collapse supernovae across a broad frequency spectrum, ranging from 50 Hz to 2 kHz where we assume the GW emission occurred when coincident data are available in the on-source window. Considering an ellipsoid model for a rotating proto-neutron star, our search is sensitive to gravitational-wave energy $1 \times 10^{-5} M_{\odot} c^2$ and luminosity $4 \times 10^{-5} M_{\odot} c^2/\text{s}$ for a source emitting at 50 Hz. These constraints are around an order of magnitude more stringent than those obtained so far with gravitational-wave data. The constraint on the ellipticity of the proto-neutron star that is formed is as low as $1.04$, at frequencies above $1200$ Hz, surpassing results from SN 2019ejj., Comment: Main paper: 6 pages, 4 figures and 1 table. Total with appendices: 20 pages, 4 figures, and 1 table

Document structure editing involves manipulating localized textual, visual, and layout components in document images based on the user's requests. Past works have shown that multimodal grounding of user requests in the document image and identifying the accurate structural components and their associated attributes remain key challenges for this task. To address these, we introduce the DocEdit-v2, a novel framework that performs end-to-end document editing by leveraging Large Multimodal Models (LMMs). It consists of three novel components: (1) Doc2Command, which simultaneously localizes edit regions of interest (RoI) and disambiguates user edit requests into edit commands; (2) LLM-based Command Reformulation prompting to tailor edit commands originally intended for specialized software into edit instructions suitable for generalist LMMs. (3) Moreover, DocEdit-v2 processes these outputs via Large Multimodal Models like GPT-4V and Gemini, to parse the document layout, execute edits on grounded Region of Interest (RoI), and generate the edited document image. Extensive experiments on the DocEdit dataset show that DocEdit-v2 significantly outperforms strong baselines on edit command generation (2-33%), RoI bounding box detection (12-31%), and overall document editing (1-12\%) tasks., Comment: EMNLP 2024 (Main)

At the forefront of control techniques is Model Predictive Control (MPC). While MPCs are effective, their requisite to recompute an optimal control given a new state leads to sparse response to the system and may make their implementation infeasible in small systems with low computational resources. To address these limitations in stability control, this research presents a small deterministic Physics-Informed MPC Surrogate model (PI-MPCS). PI-MPCS was developed to approximate the control by an MPC while encouraging stability and robustness through the integration of the system dynamics and the formation of a Lyapunov stability profile. Empirical results are presented on the task of 2D quadcopter landing. They demonstrate a rapid and precise MPC approximation on a non-linear system along with an estimated two times speed up on the computational requirements when compared against an MPC. PI-MPCS, in addition, displays a level of stable control for in- and out-of-distribution states as encouraged by the discrete dynamics residual and Lyapunov stability loss functions. PI-MPCS is meant to serve as a surrogate to MPC on situations in which the computational resources are limited.

Designing adaptive documents that are visually appealing across various devices and for diverse viewers is a challenging task. This is due to the wide variety of devices and different viewer requirements and preferences. Alterations to a document's content, style, or layout often necessitate numerous adjustments, potentially leading to a complete layout redesign. We introduce FlexDoc, a framework for creating and consuming documents that seamlessly adapt to different devices, author, and viewer preferences and interactions. It eliminates the need for manually creating multiple document layouts, as FlexDoc enables authors to define desired document properties using templates and employs both discrete and continuous optimization in a novel comprehensive optimization process, which leverages automatic text summarization and image carving techniques to adapt both layout and content during consumption dynamically. Furthermore, we demonstrate FlexDoc in multiple real-world application scenarios, such as news readers and academic papers.

As machine intelligence evolves, the need to test and compare the problem-solving abilities of different AI models grows. However, current benchmarks are often overly simplistic, allowing models to perform uniformly well, making it difficult to distinguish their capabilities. Additionally, benchmarks typically rely on static question-answer pairs, which models might memorize or guess. To address these limitations, we introduce the Dynamic Intelligence Assessment (DIA), a novel methodology for testing AI models using dynamic question templates and improved metrics across multiple disciplines such as mathematics, cryptography, cybersecurity, and computer science. The accompanying DIA-Bench dataset, which includes 150 diverse and challenging task templates with mutable parameters, is presented in various formats such as text, PDFs, compiled binaries, and visual puzzles. Our framework introduces four new metrics to assess a model's reliability and confidence across multiple attempts. These metrics revealed that even simple questions are frequently answered incorrectly when posed in varying forms, highlighting significant gaps in models' reliability. Notably, models like GPT-4o tended to overestimate their mathematical abilities, while ChatGPT-4o demonstrated better decision-making and performance through effective tool usage. We evaluated eight state-of-the-art large language models (LLMs) using DIA-Bench, showing that current models struggle with complex tasks and often display unexpectedly low confidence, even with simpler questions. The DIA framework sets a new standard for assessing not only problem-solving but also a model's adaptive intelligence and ability to assess its own limitations. The dataset is publicly available on our project's website.

As Large Language Models (LLMs) become increasingly integrated into various facets of society, a significant portion of online text consequently become synthetic. This raises concerns about bias amplification, a phenomenon where models trained on synthetic data amplify the pre-existing biases over successive training iterations. Previous literature seldom discusses bias amplification as an independent issue from model collapse. In this work, we address the gap in understanding the bias amplification of LLMs with four main contributions. Firstly, we propose a theoretical framework, defining the necessary and sufficient conditions for its occurrence, and emphasizing that it occurs independently of model collapse. Using statistical simulations with weighted maximum likelihood estimation, we demonstrate the framework and show how bias amplification arises without the sampling and functional form issues that typically drive model collapse. Secondly, we conduct experiments with GPT-2 to empirically demonstrate bias amplification, specifically examining open-ended generational political bias with a benchmark we developed. We observe that GPT-2 exhibits a right-leaning bias in sentence continuation tasks and that the bias progressively increases with iterative fine-tuning on synthetic data generated by previous iterations. Thirdly, we explore three potential mitigation strategies: Overfitting, Preservation, and Accumulation. We find that both Preservation and Accumulation effectively mitigate bias amplification and model collapse. Finally, using novel mechanistic interpretation techniques, we demonstrate that in the GPT-2 experiments, bias amplification and model collapse are driven by distinct sets of neurons, which aligns with our theoretical framework., Comment: Submitted to ARR Roling Review October

As trajectories sampled by policies used by reinforcement learning (RL) and generative flow networks (GFlowNets) grow longer, credit assignment and exploration become more challenging, and the long planning horizon hinders mode discovery and generalization. The challenge is particularly pronounced in entropy-seeking RL methods, such as generative flow networks, where the agent must learn to sample from a structured distribution and discover multiple high-reward states, each of which take many steps to reach. To tackle this challenge, we propose an approach to incorporate the discovery of action abstractions, or high-level actions, into the policy optimization process. Our approach involves iteratively extracting action subsequences commonly used across many high-reward trajectories and `chunking' them into a single action that is added to the action space. In empirical evaluation on synthetic and real-world environments, our approach demonstrates improved sample efficiency performance in discovering diverse high-reward objects, especially on harder exploration problems. We also observe that the abstracted high-order actions are interpretable, capturing the latent structure of the reward landscape of the action space. This work provides a cognitively motivated approach to action abstraction in RL and is the first demonstration of hierarchical planning in amortized sequential sampling.

A systematic investigation of superheavy nuclei in the isotopic chains of proton numbers Z=106, 114, 120, and 126 together with isotonic chains of neutron numbers N=162, 172, and 184 is presented in the theoretical framework of relativistic mean-field density functionals based on density-dependent meson-nucleon couplings. Ground-state properties, including binding energy, shape, deformation, density profile, and radius, are estimated to provide compelling evidence of magicity in these even-even nuclei, aligning with the concept of the 'island of stability'. The analysis reveals central depletion in the charge density, indicating a bubble-like structure, primarily attributed to the substantial repulsive Coulomb field and the influence of higher l-states. A thorough examination of potential decay modes, employing various semi-empirical formulas, is presented. The probable alpha-decay chains are evaluated, demonstrating excellent agreement with available experimental data, Comment: 24 Pages, 8 figures, 4 Tables, Accepted in Nuclear Physics A

Robust content moderation classifiers are essential for the safety of Generative AI systems. Content moderation, or safety classification, is notoriously ambiguous: differences between safe and unsafe inputs are often extremely subtle, making it difficult for classifiers (and indeed, even humans) to properly distinguish violating vs. benign samples without further context or explanation. Furthermore, as these technologies are deployed across various applications and audiences, scaling risk discovery and mitigation through continuous model fine-tuning becomes increasingly challenging and costly. To address these challenges, we propose a Classification approach employing Retrieval-Augmented Generation (Class-RAG). Class-RAG extends the capability of its base LLM through access to a retrieval library which can be dynamically updated to enable semantic hotfixing for immediate, flexible risk mitigation. Compared to traditional fine-tuned models, Class-RAG demonstrates flexibility and transparency in decision-making. As evidenced by empirical studies, Class-RAG outperforms on classification and is more robust against adversarial attack. Besides, our findings suggest that Class-RAG performance scales with retrieval library size, indicating that increasing the library size is a viable and low-cost approach to improve content moderation., Comment: 11 pages, submit to ACL

Reasoning and linguistic skills form the cornerstone of human intelligence, facilitating problem-solving and decision-making. Recent advances in Large Language Models (LLMs) have led to impressive linguistic capabilities and emergent reasoning behaviors, fueling widespread adoption across application domains. However, LLMs still struggle with complex reasoning tasks, highlighting their systemic limitations. In this work, we focus on evaluating whether LLMs have the requisite representations to reason using two foundational relationships: "equivalence" and "inheritance". We introduce novel tasks and benchmarks spanning six languages and observe that current SOTA LLMs often produce conflicting answers to the same questions across languages in 17.3-57.5% of cases and violate inheritance constraints in up to 37.2% cases. To enhance consistency across languages, we propose novel "Compositional Representations" where tokens are represented as composition of equivalent tokens across languages, with resulting conflict reduction (up to -4.7%) indicating benefits of shared LLM representations.

Perceiving and generating diverse modalities are crucial for AI models to effectively learn from and engage with real-world signals, necessitating reliable evaluations for their development. We identify two major issues in current evaluations: (1) inconsistent standards, shaped by different communities with varying protocols and maturity levels; and (2) significant query, grading, and generalization biases. To address these, we introduce MixEval-X, the first any-to-any, real-world benchmark designed to optimize and standardize evaluations across diverse input and output modalities. We propose multi-modal benchmark mixture and adaptation-rectification pipelines to reconstruct real-world task distributions, ensuring evaluations generalize effectively to real-world use cases. Extensive meta-evaluations show our approach effectively aligns benchmark samples with real-world task distributions. Meanwhile, MixEval-X's model rankings correlate strongly with that of crowd-sourced real-world evaluations (up to 0.98) while being much more efficient. We provide comprehensive leaderboards to rerank existing models and organizations and offer insights to enhance understanding of multi-modal evaluations and inform future research.

Establishing secure data communication necessitates secure key exchange over a public channel. Quantum key distribution (QKD), which leverages the principles of quantum physics, can achieve this with information-theoretic security. The discrete modulated (DM) continuous variable (CV) QKD protocol, in particular, is a suitable candidate for large-scale deployment of quantum-safe communication due to its simplicity and compatibility with standard high-speed telecommunication technology. Here, we present the first experimental demonstration of a four-state DM CVQKD system, successfully generating composable finite-size keys, secure against collective attacks over a 20 km fiber channel with 2.3 \times 10^{9} coherent quantum states, achieving a positive composable key rate of 11.04 \times 10^{-3} bits/symbol. This accomplishment is enabled by using an advanced security proof, meticulously selecting its parameters, and the fast, stable operation of the system. Our results mark a significant step toward the large-scale deployment of practical, high-performance, cost-effective, and highly secure quantum key distribution networks using standard telecommunication components.

Assessing the capabilities and limitations of large language models (LLMs) has garnered significant interest, yet the evaluation of multiple models in real-world scenarios remains rare. Multilingual evaluation often relies on translated benchmarks, which typically do not capture linguistic and cultural nuances present in the source language. This study provides an extensive assessment of 24 LLMs on real world data collected from Indian patients interacting with a medical chatbot in Indian English and 4 other Indic languages. We employ a uniform Retrieval Augmented Generation framework to generate responses, which are evaluated using both automated techniques and human evaluators on four specific metrics relevant to our application. We find that models vary significantly in their performance and that instruction tuned Indic models do not always perform well on Indic language queries. Further, we empirically show that factual correctness is generally lower for responses to Indic queries compared to English queries. Finally, our qualitative work shows that code-mixed and culturally relevant queries in our dataset pose challenges to evaluated models., Comment: Under Review

Reasoning is a fundamental substrate for solving novel and complex problems. Deliberate efforts in learning and developing frameworks around System 2 reasoning have made great strides, yet problems of sufficient complexity remain largely out of reach for open models. To address this gap, we examine the potential of Generative Flow Networks as a fine-tuning method for LLMs to unlock advanced reasoning capabilities. In this paper, we present a proof of concept in the domain of formal reasoning, specifically in the Neural Theorem Proving (NTP) setting, where proofs specified in a formal language such as Lean can be deterministically and objectively verified. Unlike classical reward-maximization reinforcement learning, which frequently over-exploits high-reward actions and fails to effectively explore the state space, GFlowNets have emerged as a promising approach for sampling compositional objects, improving generalization, and enabling models to maintain diverse hypotheses. Our early results demonstrate GFlowNet fine-tuning's potential for enhancing model performance in a search setting, which is especially relevant given the paradigm shift towards inference time compute scaling and "thinking slowly."

In this work, we introduce SeQuiFi, a novel approach for mitigating catastrophic forgetting (CF) in speech emotion recognition (SER). SeQuiFi adopts a sequential class-finetuning strategy, where the model is fine-tuned incrementally on one emotion class at a time, preserving and enhancing retention for each class. While various state-of-the-art (SOTA) methods, such as regularization-based, memory-based, and weight-averaging techniques, have been proposed to address CF, it still remains a challenge, particularly with diverse and multilingual datasets. Through extensive experiments, we demonstrate that SeQuiFi significantly outperforms both vanilla fine-tuning and SOTA continual learning techniques in terms of accuracy and F1 scores on multiple benchmark SER datasets, including CREMA-D, RAVDESS, Emo-DB, MESD, and SHEMO, covering different languages.

Auto-regressive Large Language Models (LLMs) demonstrate remarkable performance across domanins such as vision and language processing. However, due to sequential processing through a stack of transformer layers, autoregressive decoding faces significant computation/latency challenges, particularly in resource constrained environments like mobile and edge devices. Existing approaches in literature that aim to improve latency via skipping layers have two distinct flavors - 1) Early exit 2) Input-agnostic heuristics where tokens exit at pre-determined layers irrespective of input sequence. Both the above strategies have limitations - the former cannot be applied to handle KV Caching necessary for speed-ups in modern framework and the latter does not capture the variation in layer importance across tasks or more generally, across input sequences. To address both limitations, we propose FIRST, an algorithm that reduces inference latency by using layer-specific routers to select a subset of transformer layers adaptively for each input sequence - the prompt (during prefill stage) decides which layers will be skipped during decoding. FIRST preserves compatibility with KV caching enabling faster inference while being quality-aware. FIRST is model-agnostic and can be easily enabled on any pre-trained LLM. We further improve performance by incorporating LoRA adapters for fine-tuning on external datasets, enhancing task-specific accuracy while maintaining latency benefits. Our approach reveals that input adaptivity is critical - indeed, different task-specific middle layers play a crucial role in evolving hidden representations depending on task. Extensive experiments show that FIRST significantly reduces latency while retaining competitive performance (as compared to baselines), making our approach an efficient solution for LLM deployment in low-resource environments., Comment: 17 pages, 6 figures, Submitted to ICLR 2025

Electroencephalography (EEG)--based turn intention prediction for lower limb movement is important to build an efficient brain-computer interface (BCI) system. This study investigates the feasibility of intention detection of left-turn, right-turn, and straight walk by utilizing EEG signals obtained before the event occurrence. Synchronous data was collected using 31-channel EEG and IMU-based motion capture systems for nine healthy participants while performing left-turn, right-turn, and straight walk movements. EEG data was preprocessed with steps including Artifact Subspace Reconstruction (ASR), re-referencing, and Independent Component Analysis (ICA) to remove data noise. Feature extraction from the preprocessed EEG data involved computing various statistical measures (mean, median, standard deviation, skew, and kurtosis), and Hjorth parameters (activity, mobility, and complexity). Further, the feature selection was performed using the Random forest algorithm for the dimensionality reduction. The feature set obtained was utilized for 3-class classification using XG boost, gradient boosting, and support vector machine (SVM) with RBF kernel classifiers in a five-fold cross-validation scheme. Using the proposed intention detection methodology, the SVM classifier using an EEG window of 1.5 s and 0 s time-lag has the best decoding performance with mean accuracy, precision, and recall of 81.23%, 85.35%, and 83.92%, respectively, across the nine participants. The decoding analysis shows the feasibility of turn intention prediction for lower limb movement using the EEG signal before the event onset., Comment: *Equal contributions by the authors

We carried out a comprehensive crystal structure characterization of Ti-doped lithium ruthenate (Li$_2$Ti$_x$Ru$_{1-x}$O$_3$), to investigate the effect of Ti-doping on the structural phase transition. Experimental tools sensitive to the average structure (X-ray diffraction), as well as those sensitive to local structure (Extended X-ray Absorption Fine Structure, EXAFS; pair distribution function, PDF) are used. We observed non-monotonic dependence of the structural transition temperature on the Ti-doping level. At low doping, the transition temperature slightly increases with doping, while at high doping, the temperature decreases significantly with doping. We note two important observations from our studies. First, Ti K-edge EXAFS data shows persistent Ti-Ru dimerization even with substantial Ti doping. Second, we were able to use the PDF data to estimate the dimer correlation length above the transition temperature, which would correspond to the size of the proposed local `dimer clusters' formed by Ru-Ru and Ti-Ru neighbours. The dimer correlation length is found to be around 10~\AA, which remains robust regardless of doping. Our study therefore suggests that Ti$^{4+}$ with its $d^0$ electronic configuration is a special type of dopant when replacing Ru., Comment: 12 pages, 11 figures

In reinforcement learning, off-policy actor-critic approaches like DDPG and TD3 are based on the deterministic policy gradient. Herein, the Q-function is trained from off-policy environment data and the actor (policy) is trained to maximize the Q-function via gradient ascent. We observe that in complex tasks like dexterous manipulation and restricted locomotion, the Q-value is a complex function of action, having several local optima or discontinuities. This poses a challenge for gradient ascent to traverse and makes the actor prone to get stuck at local optima. To address this, we introduce a new actor architecture that combines two simple insights: (i) use multiple actors and evaluate the Q-value maximizing action, and (ii) learn surrogates to the Q-function that are simpler to optimize with gradient-based methods. We evaluate tasks such as restricted locomotion, dexterous manipulation, and large discrete-action space recommender systems and show that our actor finds optimal actions more frequently and outperforms alternate actor architectures.

Edge computing has emerged as an alternative to reduce transmission and processing delay and preserve privacy of the video streams. However, the ever-increasing complexity of Deep Neural Networks (DNNs) used in video-based applications (e.g. object detection) exerts pressure on memory-constrained edge devices. Model merging is proposed to reduce the DNNs' memory footprint by keeping only one copy of merged layers' weights in memory. In existing model merging techniques, (i) only architecturally identical layers can be shared; (ii) requires computationally expensive retraining in the cloud; (iii) assumes the availability of ground truth for retraining. The re-evaluation of a merged model's performance, however, requires a validation dataset with ground truth, typically runs at the cloud. Common metrics to guide the selection of shared layers include the size or computational cost of shared layers or representation size. We propose a new model merging scheme by sharing representations (i.e., outputs of layers) at the edge, guided by representation similarity S. We show that S is extremely highly correlated with merged model's accuracy with Pearson Correlation Coefficient |r| > 0.94 than other metrics, demonstrating that representation similarity can serve as a strong validation accuracy indicator without ground truth. We present our preliminary results of the newly proposed model merging scheme with identified challenges, demonstrating a promising research future direction., Comment: 3 pages, 4 figures, ACM MobiCom '24, November 18-22, 2024, Washington D.C., DC, USA

Large Language Models (LLMs) are being adopted across a wide range of tasks, including decision-making processes in industries where bias in AI systems is a significant concern. Recent research indicates that LLMs can harbor implicit biases even when they pass explicit bias evaluations. Building upon the frameworks of the LLM Implicit Association Test (IAT) Bias and LLM Decision Bias, this study highlights that newer or larger language models do not automatically exhibit reduced bias; in some cases, they displayed higher bias scores than their predecessors, such as in Meta's Llama series and OpenAI's GPT models. This suggests that increasing model complexity without deliberate bias mitigation strategies can unintentionally amplify existing biases. The variability in bias scores within and across providers underscores the need for standardized evaluation metrics and benchmarks for bias assessment. The lack of consistency indicates that bias mitigation is not yet a universally prioritized goal in model development, which can lead to unfair or discriminatory outcomes. By broadening the detection of implicit bias, this research provides a more comprehensive understanding of the biases present in advanced models and underscores the critical importance of addressing these issues to ensure the development of fair and responsible AI systems.

In recent years, there has been a significant rise in the phenomenon of hate against women on social media platforms, particularly through the use of misogynous memes. These memes often target women with subtle and obscure cues, making their detection a challenging task for automated systems. Recently, Large Language Models (LLMs) have shown promising results in reasoning using Chain-of-Thought (CoT) prompting to generate the intermediate reasoning chains as the rationale to facilitate multimodal tasks, but often neglect cultural diversity and key aspects like emotion and contextual knowledge hidden in the visual modalities. To address this gap, we introduce a Multimodal Multi-hop CoT (M3Hop-CoT) framework for Misogynous meme identification, combining a CLIP-based classifier and a multimodal CoT module with entity-object-relationship integration. M3Hop-CoT employs a three-step multimodal prompting principle to induce emotions, target awareness, and contextual knowledge for meme analysis. Our empirical evaluation, including both qualitative and quantitative analysis, validates the efficacy of the M3Hop-CoT framework on the SemEval-2022 Task 5 (MAMI task) dataset, highlighting its strong performance in the macro-F1 score. Furthermore, we evaluate the model's generalizability by evaluating it on various benchmark meme datasets, offering a thorough insight into the effectiveness of our approach across different datasets., Comment: 34 Pages. Accepted in The 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP 2024). Main Conference