In the recent past, a popular way of evaluating natural language understanding (NLU), was to consider a model's ability to perform natural language inference (NLI) tasks. In this paper, we investigate if NLI tasks, that are rarely used for LLM evaluation, can still be informative for evaluating LLMs. Focusing on five different NLI benchmarks across six models of different scales, we investigate if they are able to discriminate models of different size and quality and how their accuracies develop during training. Furthermore, we investigate the extent to which the softmax distributions of models align with human distributions in cases where statements are ambiguous or vague. Overall, our results paint a positive picture for the NLI tasks: we find that they are able to discriminate well between models at various stages of training, yet are not (all) saturated. Furthermore, we find that while the similarity of model distributions with human label distributions increases with scale, it is still much higher than the similarity between two populations of humans, making it a potentially interesting statistic to consider.

This paper explores the potential of contextualized word embeddings (CWEs) as a new tool in the history, philosophy, and sociology of science (HPSS) for studying contextual and evolving meanings of scientific concepts. Using the term "Planck" as a test case, I evaluate five BERT-based models with varying degrees of domain-specific pretraining, including my custom model Astro-HEP-BERT, trained on the Astro-HEP Corpus, a dataset containing 21.84 million paragraphs from 600,000 articles in astrophysics and high-energy physics. For this analysis, I compiled two labeled datasets: (1) the Astro-HEP-Planck Corpus, consisting of 2,900 labeled occurrences of "Planck" sampled from 1,500 paragraphs in the Astro-HEP Corpus, and (2) a physics-related Wikipedia dataset comprising 1,186 labeled occurrences of "Planck" across 885 paragraphs. Results demonstrate that the domain-adapted models outperform the general-purpose ones in disambiguating the target term, predicting its known meanings, and generating high-quality sense clusters, as measured by a novel purity indicator I developed. Additionally, this approach reveals semantic shifts in the target term over three decades in the unlabeled Astro-HEP Corpus, highlighting the emergence of the Planck space mission as a dominant sense. The study underscores the importance of domain-specific pretraining for analyzing scientific language and demonstrates the cost-effectiveness of adapting pretrained models for HPSS research. By offering a scalable and transferable method for modeling the meanings of scientific concepts, CWEs open up new avenues for investigating the socio-historical dynamics of scientific discourses.

Large Multimodal Models (LMMs) have demonstrated remarkable capabilities. While existing benchmarks for evaluating LMMs mainly focus on image comprehension, few works evaluate them from the image generation perspective. To address this issue, we propose a straightforward automated evaluation pipeline. Specifically, this pipeline requires LMMs to generate an image-prompt from a given input image. Subsequently, it employs text-to-image generative models to create a new image based on these generated prompts. Finally, we evaluate the performance of LMMs by comparing the original image with the generated one. Furthermore, we introduce MMGenBench-Test, a comprehensive benchmark developed to evaluate LMMs across 13 distinct image patterns, and MMGenBench-Domain, targeting the performance evaluation of LMMs within the generative image domain. A thorough evaluation involving over 50 popular LMMs demonstrates the effectiveness and reliability in both the pipeline and benchmark. Our observations indicate that numerous LMMs excelling in existing benchmarks fail to adequately complete the basic tasks, related to image understanding and description. This finding highlights the substantial potential for performance improvement in current LMMs and suggests avenues for future model optimization. Concurrently, our pipeline facilitates the efficient assessment of LMMs performance across diverse domains by using solely image inputs.

Large language models (LLMs) deliver impressive results but face challenges from increasing model sizes and computational costs. Structured pruning reduces model size and speeds up inference but often causes uneven degradation across domains, leading to biased performance. To address this, we propose DRPruning, which incorporates distributionally robust optimization to restore balanced performance across domains, along with further improvements to enhance robustness. Experiments in monolingual and multilingual settings show that our method surpasses similarly sized models in pruning and continued pretraining over perplexity, downstream tasks, and instruction tuning. We further provide analysis demonstrating the robustness of our method towards various domains and distribution shifts. Furthermore, our method automatically determines optimal reference losses and data ratios, suggesting potential for broader applications. Our code is available at https://github.com/hexuandeng/DRPruning.

This study investigates language models' generative capabilities in tool-use dialogs. We categorize the models' outputs in tool-use dialogs into four distinct types: Tool Call, Answer Completion, Slot Question, and Relevance Detection, which serve as aspects for evaluation. We introduce FunctionChat-Bench, comprising 700 evaluation items and automated assessment programs. Using this benchmark, we evaluate several language models that support function calling. Our findings indicate that while language models may exhibit high accuracy in single-turn Tool Call scenarios, this does not necessarily translate to superior generative performance in multi-turn environments. We argue that the capabilities required for function calling extend beyond generating tool call messages; they must also effectively generate conversational messages that engage the user.

Predicting future international events from textual information, such as news articles, has tremendous potential for applications in global policy, strategic decision-making, and geopolitics. However, existing datasets available for this task are often limited in quality, hindering the progress of related research. In this paper, we introduce WORLDREP (WORLD Relationship and Event Prediction), a novel dataset designed to address these limitations by leveraging the advanced reasoning capabilities of large-language models (LLMs). Our dataset features high-quality scoring labels generated through advanced prompt modeling and rigorously validated by domain experts in political science. We showcase the quality and utility of WORLDREP for real-world event prediction tasks, demonstrating its effectiveness through extensive experiments and analysis. Furthermore, we publicly release our dataset along with the full automation source code for data collection, labeling, and benchmarking, aiming to support and advance research in text-based event prediction.

The emergence of generative artificial intelligence (GAI) and large language models (LLMs) such ChatGPT has enabled the realization of long-harbored desires in software and robotic development. The technology however, has brought with it novel ethical challenges. These challenges are compounded by the application of LLMs in other machine learning systems, such as multi-robot systems. The objectives of the study were to examine novel ethical issues arising from the application of LLMs in multi-robot systems. Unfolding ethical issues in GPT agent behavior (deliberation of ethical concerns) was observed, and GPT output was compared with human experts. The article also advances a model for ethical development of multi-robot systems. A qualitative workshop-based method was employed in three workshops for the collection of ethical concerns: two human expert workshops (N=16 participants) and one GPT-agent-based workshop (N=7 agents; two teams of 6 agents plus one judge). Thematic analysis was used to analyze the qualitative data. The results reveal differences between the human-produced and GPT-based ethical concerns. Human experts placed greater emphasis on new themes related to deviance, data privacy, bias and unethical corporate conduct. GPT agents emphasized concerns present in existing AI ethics guidelines. The study contributes to a growing body of knowledge in context-specific AI ethics and GPT application. It demonstrates the gap between human expert thinking and LLM output, while emphasizing new ethical concerns emerging in novel technology.

Visual instruction tuning (VIT) enables multimodal large language models (MLLMs) to effectively handle a wide range of vision tasks by framing them as language-based instructions. Building on this, continual visual instruction tuning (CVIT) extends the capability of MLLMs to incrementally learn new tasks, accommodating evolving functionalities. While prior work has advanced CVIT through the development of new benchmarks and approaches to mitigate catastrophic forgetting, these efforts largely follow traditional continual learning paradigms, neglecting the unique challenges specific to CVIT. We identify a dual form of catastrophic forgetting in CVIT, where MLLMs not only forget previously learned visual understanding but also experience a decline in instruction following abilities as they acquire new tasks. To address this, we introduce the Separable Mixture of Low-Rank Adaptation (SMoLoRA) framework, which employs separable routing through two distinct modules - one for visual understanding and another for instruction following. This dual-routing design enables specialized adaptation in both domains, preventing forgetting while improving performance. Furthermore, we propose a novel CVIT benchmark that goes beyond existing benchmarks by additionally evaluating a model's ability to generalize to unseen tasks and handle diverse instructions across various tasks. Extensive experiments demonstrate that SMoLoRA outperforms existing methods in mitigating dual forgetting, improving generalization to unseen tasks, and ensuring robustness in following diverse instructions.

Reproducing buggy code is the first and crucially important step in issue resolving, as it aids in identifying the underlying problems and validating that generated patches resolve the problem. While numerous approaches have been proposed for this task, they primarily address common, widespread errors and struggle to adapt to unique, evolving errors specific to individual code repositories. To fill this gap, we propose EvoCoder, a multi-agent continuous learning framework for issue code reproduction. EvoCoder adopts a reflection mechanism that allows the LLM to continuously learn from previously resolved problems and dynamically refine its strategies to new emerging challenges. To prevent experience bloating, EvoCoder introduces a novel hierarchical experience pool that enables the model to adaptively update common and repo-specific experiences. Our experimental results show a 20\% improvement in issue reproduction rates over existing SOTA methods. Furthermore, integrating our reproduction mechanism significantly boosts the overall accuracy of the existing issue-resolving pipeline.

How are LLM-based agents used in the future? While many of the existing work on agents has focused on improving the performance of a specific family of objective and challenging tasks, in this work, we take a different perspective by thinking about full delegation: agents take over humans' routine decision-making processes and are trusted by humans to find solutions that fit people's personalized needs and are adaptive to ever-changing context. In order to achieve such a goal, the behavior of the agents, i.e., agentic behaviors, should be evaluated not only on their achievements (i.e., outcome evaluation), but also how they achieved that (i.e., procedure evaluation). For this, we propose APEC Agent Constitution, a list of criteria that an agent should follow for good agentic behaviors, including Accuracy, Proactivity, Efficiency and Credibility. To verify whether APEC aligns with human preferences, we develop APEC-Travel, a travel planning agent that proactively extracts hidden personalized needs via multi-round dialog with travelers. APEC-Travel is constructed purely from synthetic data generated by Llama3.1-405B-Instruct with a diverse set of travelers' persona to simulate rich distribution of dialogs. Iteratively fine-tuned to follow APEC Agent Constitution, APEC-Travel surpasses baselines by 20.7% on rule-based metrics and 9.1% on LLM-as-a-Judge scores across the constitution axes.