Search

Large Language Models (LLMs) have demonstrated remarkable performance across a wide range of natural language processing tasks. However, they are often distracted by irrelevant or noisy context in input sequences that degrades output quality. This problem affects both long- and short-context scenarios, such as retrieval-augmented generation, table question-answering, and in-context learning. We reveal that LLMs can implicitly identify whether input sequences contain useful information at early layers, prior to token generation. Leveraging this insight, we introduce Early Noise Dropping (\textsc{END}), a novel approach to mitigate this issue without requiring fine-tuning the LLMs. \textsc{END} segments input sequences into chunks and employs a linear prober on the early layers of LLMs to differentiate between informative and noisy chunks. By discarding noisy chunks early in the process, \textsc{END} preserves critical information, reduces distraction, and lowers computational overhead. Extensive experiments demonstrate that \textsc{END} significantly improves both performance and efficiency across different LLMs on multiple evaluation datasets. Furthermore, by investigating LLMs' implicit understanding to the input with the prober, this work also deepens understanding of how LLMs do reasoning with contexts internally.

The collaborative design process is intrinsically complicated and dynamic, and researchers have long been exploring how to enhance efficiency in this process. As Artificial Intelligence technology evolves, it has been widely used as a design tool and exhibited the potential as a design collaborator. Nevertheless, problems concerning how designers should communicate with AI in collaborative design remain unsolved. To address this research gap, we referred to how designers communicate fluently in human-human design collaboration, and found awareness to be an important ability for facilitating communication by understanding their collaborators and current situation. However, previous research mainly studied and supported human awareness, the possible impact AI awareness would bring to the human-AI collaborative design process, and the way to realize AI awareness remain unknown. In this study, we explored how AI awareness will impact human-AI collaboration through a Wizard-of-Oz experiment. Both quantitative and qualitative results supported that enabling AI to have awareness can enhance the communication fluidity between human and AI, thus enhancing collaboration efficiency. We further discussed the results and concluded design implications for future human-AI collaborative design systems., Comment: 16 pages, 4 figures, accepted by IUI 2025

Due to the scarcity of agent-oriented pre-training data, LLM-based autonomous agents typically rely on complex prompting or extensive fine-tuning, which often fails to introduce new capabilities while preserving strong generalizability. We introduce Hephaestus-Forge, the first large-scale pre-training corpus designed to enhance the fundamental capabilities of LLM agents in API function calling, intrinsic reasoning and planning, and adapting to environmental feedback. Hephaestus-Forge comprises 103B agent-specific data encompassing 76,537 APIs, including both tool documentation to introduce knowledge of API functions and function calling trajectories to strengthen intrinsic reasoning. To explore effective training protocols, we investigate scaling laws to identify the optimal recipe in data mixing ratios. By continual pre-training on Hephaestus-Forge, Hephaestus outperforms small- to medium-scale open-source LLMs and rivals commercial LLMs on three agent benchmarks, demonstrating the effectiveness of our pre-training corpus in enhancing fundamental agentic capabilities and generalization of LLMs to new tasks or environments., Comment: Accepted to NAACL 2025 main conference

With the rise of generative large language models (LLMs) like LLaMA and ChatGPT, these models have significantly transformed daily life and work by providing advanced insights. However, as jailbreak attacks continue to circumvent built-in safety mechanisms, exploiting carefully crafted scenarios or tokens, the safety risks of LLMs have come into focus. While numerous defense strategies--such as prompt detection, modification, and model fine-tuning--have been proposed to counter these attacks, a critical question arises: do these defenses compromise the utility and usability of LLMs for legitimate users? Existing research predominantly focuses on the effectiveness of defense strategies without thoroughly examining their impact on performance, leaving a gap in understanding the trade-offs between LLM safety and performance. Our research addresses this gap by conducting a comprehensive study on the utility degradation, safety elevation, and exaggerated-safety escalation of LLMs with jailbreak defense strategies. We propose USEBench, a novel benchmark designed to evaluate these aspects, along with USEIndex, a comprehensive metric for assessing overall model performance. Through experiments on seven state-of-the-art LLMs, we found that mainstream jailbreak defenses fail to ensure both safety and performance simultaneously. Although model-finetuning performs the best overall, their effectiveness varies across LLMs. Furthermore, vertical comparisons reveal that developers commonly prioritize performance over safety when iterating or fine-tuning their LLMs.

Discovering dominant patterns and exploring dynamic behaviors especially critical state transitions and tipping points in high-dimensional time-series data are challenging tasks in study of real-world complex systems, which demand interpretable data representations to facilitate comprehension of both spatial and temporal information within the original data space. Here, we proposed a general and analytical ultralow-dimensionality reduction method for dynamical systems named spatial-temporal principal component analysis (stPCA) to fully represent the dynamics of a high-dimensional time-series by only a single latent variable without distortion, which transforms high-dimensional spatial information into one-dimensional temporal information based on nonlinear delay-embedding theory. The dynamics of this single variable is analytically solved and theoretically preserves the temporal property of original high-dimensional time-series, thereby accurately and reliably identifying the tipping point before an upcoming critical transition. Its applications to real-world datasets such as individual-specific heterogeneous ICU records demonstrated the effectiveness of stPCA, which quantitatively and robustly provides the early-warning signals of the critical/tipping state on each patient.

Composing music for video is essential yet challenging, leading to a growing interest in automating music generation for video applications. Existing approaches often struggle to achieve robust music-video correspondence and generative diversity, primarily due to inadequate feature alignment methods and insufficient datasets. In this study, we present General Video-to-Music Generation model (GVMGen), designed for generating high-related music to the video input. Our model employs hierarchical attentions to extract and align video features with music in both spatial and temporal dimensions, ensuring the preservation of pertinent features while minimizing redundancy. Remarkably, our method is versatile, capable of generating multi-style music from different video inputs, even in zero-shot scenarios. We also propose an evaluation model along with two novel objective metrics for assessing video-music alignment. Additionally, we have compiled a large-scale dataset comprising diverse types of video-music pairs. Experimental results demonstrate that GVMGen surpasses previous models in terms of music-video correspondence, generative diversity, and application universality., Comment: Accepted by the 39th AAAI Conference on Artificial Intelligence (AAAI-25)

The study of the Heisenberg-Kitaev honeycomb ferromagnets has recently drawn attention because of their rich topological properties. Topological phase transitions may arise when there exist two or more distinct topological phases, and they are often revealed by a gap-closing phenomenon. In this work, we investigate the magnonic properties of honeycomb ferromagnets exhibiting Kitaev and DMI interactions in the presence of a Heisenberg exchange and magnetocrystalline anisotropy exposed to a magnetic field. We employ the Self-Consistent Renormalization (SCR) spin wave theory to investigate the effects of magnon-magnon interactions (MMIs) and thermal fluctuations on the properties of magnons. Our findings demonstrate that the magnon system undergoes topological phase transitions driven by temperature and magnetic fields, which are attributed to MMIs. Specifically, as the temperature rises, the magnon band gap at the Dirac points closes and reopens at the critical temperature Tc , which is below the Curie temperature. By showing that the Chern numbers of the magnonic bands are distinct above and below Tc , we confirm that the gap-closing phenomenon is indeed a signature for the topological phase transitions. Furthermore, our analysis indicates that the thermal Hall conductivity in the magnonic system exhibits a sign reversal at Tc , which can serve as an experimental probe of its topological nature.

Dynamic Music Emotion Recognition (DMER) aims to predict the emotion of different moments in music, playing a crucial role in music information retrieval. The existing DMER methods struggle to capture long-term dependencies when dealing with sequence data, which limits their performance. Furthermore, these methods often overlook the influence of individual differences on emotion perception, even though everyone has their own personalized emotional perception in the real world. Motivated by these issues, we explore more effective sequence processing methods and introduce the Personalized DMER (PDMER) problem, which requires models to predict emotions that align with personalized perception. Specifically, we propose a Dual-Scale Attention-Based Meta-Learning (DSAML) method. This method fuses features from a dual-scale feature extractor and captures both short and long-term dependencies using a dual-scale attention transformer, improving the performance in traditional DMER. To achieve PDMER, we design a novel task construction strategy that divides tasks by annotators. Samples in a task are annotated by the same annotator, ensuring consistent perception. Leveraging this strategy alongside meta-learning, DSAML can predict personalized perception of emotions with just one personalized annotation sample. Our objective and subjective experiments demonstrate that our method can achieve state-of-the-art performance in both traditional DMER and PDMER., Comment: Accepted by the 39th AAAI Conference on Artificial Intelligence (AAAI-25)

As one type of blue early-type galaxies, the evolutionary history and fate of star-forming lenticular galaxies (S0s) remain elusive. We selected 134 star-forming S0s from the SDSS-IV MaNGA survey and found that they have steep and warped size-mass relations, similar to quiescent S0s and red spirals, indicating that they may have similar gas dissipation scenarios. These galaxies have a higher central stellar mass surface density than normal blue spirals. The radial profiles of $D_{\rm n}4000$ and [Mgb/Fe] show that red spirals and quiescent S0s have similar old central populations and high [Mgb/Fe] values, suggesting rapid bulge formation, though red spirals exhibit a steeper gradient possibly due to residual star formation (SF) in outer regions. In contrast, star-forming S0s exhibit profiles between quiescent S0s/red spirals and normal blue spirals, with relatively flat $D_{\rm n}4000$ and [Mgb/Fe] gradients. More long-term SF history causes normal blue spirals to have very flat $D_{\rm n}4000$ and [Mgb/Fe] profiles, and the majority of them (79 $\pm$ 5 $\%$) have S$\acute{\rm e}$rsic index $<$ 2. We also found that the halo mass of star-forming S0s resembles that of quiescent S0s/red spirals, with 82 $\pm$ 5 $\%$ exceeding the critical mass ($M_{\rm halo} = 10^{12}$$M_{\odot}$h$^{-1}$). To supplement previous H\,{\sc i} detection of star-forming S0s covered by H\,{\sc i}MaNGA, we obtained new observation for H\,{\sc i} emission from 41 star-forming S0s in our sample using the Five-hundred-meter Aperture Spherical Radio Telescope. We found that the H\,{\sc i} mass distribution of star-forming S0s matches that of normal blue spirals, although both star-forming S0s and red spirals are relatively gas-poor, resulting in varying atomic gas depletion times due to different SF levels. Based on these observational results, we discuss the possible evolutionary scenarios of star-forming S0s., Comment: 22 pages, 9 figures, 4 tables, accepted for publication in ApJ

Recently, the surge of efficient and automated 3D AI-generated content (AIGC) methods has increasingly illuminated the path of transforming human imagination into complex 3D structures. However, the automated generation of 3D content is still significantly lags in industrial application. This gap exists because 3D modeling demands high-quality assets with sharp geometry, exquisite topology, and physically based rendering (PBR), among other criteria. To narrow the disparity between generated results and artists' expectations, we introduce GraphicsDreamer, a method for creating highly usable 3D meshes from single images. To better capture the geometry and material details, we integrate the PBR lighting equation into our cross-domain diffusion model, concurrently predicting multi-view color, normal, depth images, and PBR materials. In the geometry fusion stage, we continue to enforce the PBR constraints, ensuring that the generated 3D objects possess reliable texture details, supporting realistic relighting. Furthermore, our method incorporates topology optimization and fast UV unwrapping capabilities, allowing the 3D products to be seamlessly imported into graphics engines. Extensive experiments demonstrate that our model can produce high quality 3D assets in a reasonable time cost compared to previous methods.

Image Splicing Localization (ISL) is a fundamental yet challenging task in digital forensics. Although current approaches have achieved promising performance, the edge information is insufficiently exploited, resulting in poor integrality and high false alarms. To tackle this problem, we propose a multi-scale cross-fusion and edge-supervision network for ISL. Specifically, our framework consists of three key steps: multi-scale features cross-fusion, edge mask prediction and edge-supervision localization. Firstly, we input the RGB image and its noise image into a segmentation network to learn multi-scale features, which are then aggregated via a cross-scale fusion followed by a cross-domain fusion to enhance feature representation. Secondly, we design an edge mask prediction module to effectively mine the reliable boundary artifacts. Finally, the cross-fused features and the reliable edge mask information are seamlessly integrated via an attention mechanism to incrementally supervise and facilitate model training. Extensive experiments on publicly available datasets demonstrate that our proposed method is superior to state-of-the-art schemes., Comment: 5 pages,3 figures

Traditional real-time operating systems (RTOS) often exhibit poor parallel performance, while thread monitoring in Linux-based systems presents significant challenges. To address these issues, this paper proposes a satellite flight software system design based on the Robot Operating System (ROS), leveraging ROS's built-in reliable publish-subscribe messaging mechanism for inter-application communication. Considering the complex functional requirements of modern small satellites, the design incorporates both hardware and software architecture, alongside system scheduling and error-correction mechanisms. This approach ensures efficient parallel data processing and system reliability, while also reducing the development cycle through code reuse. Comprehensive testing, including system time delay, system management, fault tolerance, and system maintenance, was conducted to validate the system's capabilities in telemetry, remote control, new feature integration, and autonomous error correction. The results demonstrate the high reliability and ease of maintenance of the satellite flight software offering a reference framework for the rapid development of high-performance small satellite operations systems.

Diffusion Models enable realistic image generation, raising the risk of misinformation and eroding public trust. Currently, detecting images generated by unseen diffusion models remains challenging due to the limited generalization capabilities of existing methods. To address this issue, we rethink the effectiveness of pre-trained models trained on large-scale, real-world images. Our findings indicate that: 1) Pre-trained models can cluster the features of real images effectively. 2) Models with pre-trained weights can approximate an optimal generalization solution at a specific training step, but it is extremely unstable. Based on these facts, we propose a simple yet effective training method called Learning on Less (LoL). LoL utilizes a random masking mechanism to constrain the model's learning of the unique patterns specific to a certain type of diffusion model, allowing it to focus on less image content. This leverages the inherent strengths of pre-trained weights while enabling a more stable approach to optimal generalization, which results in the extraction of a universal feature that differentiates various diffusion-generated images from real images. Extensive experiments on the GenImage benchmark demonstrate the remarkable generalization capability of our proposed LoL. With just 1% training data, LoL significantly outperforms the current state-of-the-art, achieving a 13.6% improvement in average ACC across images generated by eight different models.

Image Forgery Localization (IFL) technology aims to detect and locate the forged areas in an image, which is very important in the field of digital forensics. However, existing IFL methods suffer from feature degradation during training using multi-layer convolutions or the self-attention mechanism, and perform poorly in detecting small forged regions and in robustness against post-processing. To tackle these, we propose a guided and multi-scale feature aggregated network for IFL. Spectifically, in order to comprehensively learn the noise feature under different types of forgery, we develop an effective noise extraction module in a guided way. Then, we design a Feature Aggregation Module (FAM) that uses dynamic convolution to adaptively aggregate RGB and noise features over multiple scales. Moreover, we propose an Atrous Residual Pyramid Module (ARPM) to enhance features representation and capture both global and local features using different receptive fields to improve the accuracy and robustness of forgery localization. Expensive experiments on 5 public datasets have shown that our proposed model outperforms several the state-of-the-art methods, specially on small region forged image., Comment: 36 pages, 6 figures

The origins of lenticular galaxies (S0s) can be classified into two main categories: ``minor mergers" in low-density environments (LDEs) and ``faded spirals" in high-density environments (HDEs). The transitional phase in the evolution of S0s, namely, star-forming lenticular galaxies (SFS0s), can serve as an important probe for analyzing the complex processes involved in the transformation between different galaxy types and the quenching of star formation (SF). We attempt to find the impact of different environments on the global properties and spatially resolved quantities of SFS0s. We selected 71 SFS0s from the SDSS-IV MaNGA Survey, comprising 23 SFS0s in HDEs (SFS0s$\_$HE) and 48 SFS0s in LDEs (SFS0s$\_$LE). We examined the effects of the environment, by studying the global properties, concentration index, and radial profiles of the derived quantities. The varied environments of SFS0s do not lead to any significant difference in global properties (e.g., S$\acute{\rm e}$rsic index). By calculating $CI_{\rm H_{\alpha}/cont}$, we observe that different environments may cause varying concentrations of SF. Specifically, SFS0s$\_$LE, affected by external gas mergers or inflow, exhibit a more centrally concentrated SF (i.e., larger $CI_{\rm H_{\alpha}/cont}$). This trend is further supported by $CI_{\rm SFR, H_{\alpha}}$, which only considers the gas disk of the galaxy. This observation is aligned with the observed shrinking of gas disks in galaxies affected by ram-pressure stripping in HDEs. Furthermore, their $\Sigma_{\rm SFR}$ or resolved sSFR are comparable. On average, SFS0s$\_$LE display significantly higher values for both quantities. Finally, the observed D$_{\rm n}4000$ and gas-phase metallicity gradient correspond well to their assumed origins. However, we did not find a significantly lower gas-phase metallicity in SFS0s$\_$LE. Abridged, Comment: 15 pages, 7 figures, 4 tables. Accepted for publication in A&A

The rise of generative models has sparked concerns about image authenticity online, highlighting the urgent need for an effective and general detector. Recent methods leveraging the frozen pre-trained CLIP-ViT model have made great progress in deepfake detection. However, these models often rely on visual-general features directly extracted by the frozen network, which contain excessive information irrelevant to the task, resulting in limited detection performance. To address this limitation, in this paper, we propose an efficient Guided and Fused Frozen CLIP-ViT (GFF), which integrates two simple yet effective modules. The Deepfake-Specific Feature Guidance Module (DFGM) guides the frozen pre-trained model in extracting features specifically for deepfake detection, reducing irrelevant information while preserving its generalization capabilities. The Multi-Stage Fusion Module (FuseFormer) captures low-level and high-level information by fusing features extracted from each stage of the ViT. This dual-module approach significantly improves deepfake detection by fully leveraging CLIP-ViT's inherent advantages. Extensive experiments demonstrate the effectiveness and generalization ability of GFF, which achieves state-of-the-art performance with optimal results in only 5 training epochs. Even when trained on only 4 classes of ProGAN, GFF achieves nearly 99% accuracy on unseen GANs and maintains an impressive 97% accuracy on unseen diffusion models.

As a novel and promising 2D material, bismuth oxyselenide (Bi$_2$O$_2$Se) has demonstrated significant potential to overcome existing technical barriers in various electronic device applications, due to its unique physical properties like high symmetry, adjustable electronic structure, ultra-high electron mobility. However, the rapid growth of Bi$_2$O$_2$Se films down to a few atomic layers with precise control remains a significant challenge. In this work, the growth of two-dimensional (2D) Bi$_2$O$_2$Se thin films by the pulsed laser deposition (PLD) method is systematically investigated. By controlling temperature, oxygen pressure, laser energy density and laser emission frequency, we successfully prepare atomically thin and flat Bi$_2$O$_2$Se (001) thin films on the (001) surface of SrTiO3. Importantly, we provide a fundamental and unique perspective toward understanding the growth process of atomically thin and flat Bi$_2$O$_2$Se films, and the growth process can be primarily summarized into four steps: i) anisotropic non-spontaneous nucleation preferentially along the step roots; ii) monolayer Bi$_2$O$_2$Se nanosheets expanding across the surrounding area, and eventually covering the entire STO substrate step; iii) vertical growth of Bi$_2$O$_2$Se monolayer in a 2D Frank-van der Merwe (FM) epitaxial growth, and iv) with a layer-by-layer 2D FM growth mode, ultimately producing an atomically flat and epitaxially aligned thin film. Moreover, the combined results of the crystallinity quality, surface morphology and the chemical states manifest the successful PLD-growth of high-quality Bi$_2$O$_2$Se films in a controllable and fast mode.

We present the integral field spectroscopic observations of ionized gas (H$\alpha$ and [{\ion{N}{II}}]) using the PCWI, along with deep CO(2-1) observations by the $^\backprime\bar{\rm U}^\backprime\bar{\rm u}$ receiver on JCMT for AGC 102004. The velocity field of H$\alpha$ shows an anomalous distribution in the North-Western (NW) disk. The H$\alpha$ spectrum is well-fitted by two Gaussian components, and the weak Gaussian component is dominated by the anomalous H$\alpha$ in the NW disk. The Gaussian fit center of H$\alpha$ emission is offset by +24.2 km s$^{-1}$ from the systemic velocity obtained from the HI emission. We derive the gas-phase metallicity, 12+log(O/H), using [{\ion{N}{II}}]$\lambda$6583/H$\alpha$ ratio as a proxy. The mean value of 12+log(O/H) is 8.30 $\pm$ 0.19 over the whole galaxy. The metallicity in the outer disk is lower than the detection limit of 7.72, indicating the metallicity gradient exists in AGC 102004. We speculate a minor/mini-merger event could have happened to the NW disk. CO(2-1) emission has non-detection in AGC 102004, reaching a noise level of 0.33 mK smoothed to 30 km s$^{-1}$. The upper limit of molecular gas mass in AGC 102004 is 2.1 $\times$ 10$^7$ M$\odot$ with X$_{\rm CO}$ = 3.02$\times$10$^{20}$ cm$^{-2}$ (K km s$^{-1}$)$^{-1}$. The M$_{\rm H_2}$/M$^{\rm corr}_{\rm HI}$ of AGC 102004 is lower than 0.0037 and lower than that of normal galaxies., Comment: 10 pages, 9 figures, Accepted for publication in ApJ

Although Large Language Models (LLMs) have demonstrated strong ability, they are further supposed to be controlled and guided by in real-world scenarios to be safe, accurate, and intelligent. This demands the possession of capability of LLMs. However, no prior work has made a clear evaluation of the inferential rule-following capability of LLMs. Previous studies that try to evaluate the inferential rule-following capability of LLMs fail to distinguish the inferential rule-following scenarios from the instruction-following scenarios. Therefore, this paper first clarifies the concept of inferential rule-following and proposes a comprehensive benchmark, RuleBench, to evaluate a diversified range of inferential rule-following abilities. Our experimental results on a variety of LLMs show that they are still limited in following rules. Our analysis based on the evaluation results provides insights into the improvements for LLMs toward a better inferential rule-following intelligent agent. We further propose Inferential Rule-Following Tuning (IRFT). The experimental results show that through IRFT, LLMs can learn abstract rule-following abilities from purely synthetic data and then generalize to RuleBench. The data and code can be found at: https://anonymous.4open.science/r/llm-rule-following-B3E3/

This paper provides a detailed survey of synthetic data techniques. We first discuss the expected goals of using synthetic data in data augmentation, which can be divided into four parts: 1) Improving Diversity, 2) Data Balancing, 3) Addressing Domain Shift, and 4) Resolving Edge Cases. Synthesizing data are closely related to the prevailing machine learning techniques at the time, therefore, we summarize the domain of synthetic data techniques into four categories: 1) Expert-knowledge, 2) Direct Training, 3) Pre-train then Fine-tune, and 4) Foundation Models without Fine-tuning. Next, we categorize the goals of synthetic data filtering into four types for discussion: 1) Basic Quality, 2) Label Consistency, and 3) Data Distribution. In section 5 of this paper, we also discuss the future directions of synthetic data and state three direction that we believe is important: 1) focus more on quality, 2) the evaluation of synthetic data, and 3) multi-model data augmentation.

High-performance phase control units are crucial in beamforming technology, which has gained substantial attention for its ability to manipulate the wireless propagation environment, thereby enhancing capacity and coverage in communication networks. This paper presents the design and fabrication of a 3.5GHz reflective liquid-crystal (LC) phase shifter. The phase shifter is constructed using coplanar differential lines, periodically loaded with floating electrodes. The LCs in the overlapping areas act as variable capacitors, and the continuous phase shift can be adjusted by applying AC to the permittivity in these areas. Both simulation and measurement results demonstrate impressive Figures of Merit (FoM) of 101.3 degrees per dB and 85.7 degrees per dB, respectively. The grounding issues typically associated with coplanar waveguides (CPWs) on glass substrates are effectively mitigated by employing the virtual ground in a differential pair configuration. The innovative reflective-type operation minimizes the unit cell size and allows for low-cost manufacturing of phase shifter arrays and advances the practical development of beamforming technology.

This demo showcases a simple tool that utilizes AIGC technology, enabling both professional designers and regular users to easily customize clothing for their digital avatars. Customization options include changing clothing colors, textures, logos, and patterns. Compared with traditional 3D modeling processes, our approach significantly enhances efficiency and interactivity and reduces production costs.

Finetuning large language models with a variety of instruction-response pairs has enhanced their capability to understand and follow instructions. Current instruction tuning primarily relies on teacher models or human intervention to generate and refine the instructions and responses for training, which are costly, non-sustainable, and may lack diversity. In this paper, we introduce Mosaic Instruction Tuning (Mosaic-IT), a human/model-free compositional data augmentation method that can efficiently create rich and diverse augmentations from existing instruction tuning data to enhance the LLMs. Mosaic-IT randomly concatenates multiple instruction data into one and trains the model to produce the corresponding responses with predefined higher-level meta-instructions to strengthen its multi-step instruction-following and format-following skills. Our extensive evaluations demonstrate a superior performance and training efficiency of Mosaic-IT, which achieves consistent performance improvements over various benchmarks and a $80\%$ reduction in training costs compared with original instruction tuning. Our codes and data are available at https://github.com/tianyi-lab/Mosaic-IT.

Large Language Models (LLMs) have shown great ability in solving traditional natural language tasks and elementary reasoning tasks with appropriate prompting techniques. However, their ability is still limited in solving complicated science problems. In this work, we aim to push the upper bound of the reasoning capability of LLMs by proposing a collaborative multi-agent, multi-reasoning-path (CoMM) prompting framework. Specifically, we prompt LLMs to play different roles in a problem-solving team, and encourage different role-play agents to collaboratively solve the target task. In particular, we discover that applying different reasoning paths for different roles is an effective strategy to implement few-shot prompting approaches in the multi-agent scenarios. Empirical results demonstrate the effectiveness of the proposed methods on two college-level science problems over competitive baselines. Our further analysis shows the necessity of prompting LLMs to play different roles or experts independently. We release the code at: https://github.com/amazon-science/comm-prompt, Comment: Accepted to NAACL 2024

The characterization of Kirkwood-Dirac (KD) classicality or non-classicality is very important in quantum information processing. In general, the set of KD classical states with respect to two bases is not a convex polytope[J. Math. Phys. \textbf{65} 072201 (2024)], which makes us interested in finding out in which circumnstances they do form a polytope. In this paper, we focus on the characterization of KD classicality of mixed states for the case where the transition matrix between two bases is a discrete Fourier transform (DFT) matrix in Hilbert space with dimensions $p^2$ and $pq$, respectively, where $p, q$ are prime. For the two particular cases we investigate, the sets of extremal points are finite, implying that the set of KD classical states we characterize forms a convex polytope. We show that for $p^2$ dimensional system, the set $\rm{KD}_{\mathcal{A},\mathcal{B}}^+$ is a convex hull of the set $\rm {pure}({\rm {KD}_{\mathcal{A},\mathcal{B}}^+})$ based on DFT, where $\rm{KD}_{\mathcal{A},\mathcal{B}}^+$ is the set of KD classical states with respect to two bases and $\rm {pure}({\rm {KD}_{\mathcal{A},\mathcal{B}}^+})$ is the set of all the rank-one projectors of KD classical pure states with respect to two bases. In $pq$ dimensional system, we believe that this result also holds. Unfortunately, we do not completely prove it, but some meaningful conclusions are obtained about the characterization of KD classicality., Comment: 26 pages