Large language models that use retrieval augmented generation have the potential to unlock valuable knowledge for researchers, policymakers, and the public by making long and technical climate-related documents more accessible. While this approach can help alleviate factual hallucinations by relying on retrieved passages as additional context, its effectiveness depends on whether the model's output remains faithful to these passages. To address this, we explore the automatic assessment of faithfulness of different models in this setting. We then focus on ClimateGPT, a large language model specialised in climate science, to examine which factors in its instruction fine-tuning impact the model's faithfulness. By excluding unfaithful subsets of the model's training data, we develop ClimateGPT Faithful+, which achieves an improvement in faithfulness from 30% to 57% in supported atomic claims according to our automatic metric.
The evaluation of large language models (LLMs) via benchmarks is widespread, yet inconsistencies between different leaderboards and poor separability among top models raise concerns about their ability to accurately reflect authentic model capabilities. This paper provides a critical analysis of benchmark effectiveness, examining main-stream prominent LLM benchmarks using results from diverse models. We first propose a new framework for accurate and reliable estimations of item characteristics and model abilities. Specifically, we propose Pseudo-Siamese Network for Item Response Theory (PSN-IRT), an enhanced Item Response Theory framework that incorporates a rich set of item parameters within an IRT-grounded architecture. Based on PSN-IRT, we conduct extensive analysis which reveals significant and varied shortcomings in the measurement quality of current benchmarks. Furthermore, we demonstrate that leveraging PSN-IRT is able to construct smaller benchmarks while maintaining stronger alignment with human preference.
Federated Learning (FL) enables distributed model training across edge devices in a privacy-friendly manner. However, its efficiency heavily depends on effective device selection and high-dimensional resource allocation in dynamic and heterogeneous wireless environments. Conventional methods demand a confluence of domain-specific expertise, extensive hyperparameter tuning, and/or heavy interaction cost. This paper proposes a Tool-aided Evolutionary Large Language Model (T-ELLM) framework to generate a qualified policy for device selection in a wireless FL environment. Unlike conventional optimization methods, T-ELLM leverages natural language-based scenario prompts to enhance generalization across varying network conditions. The framework decouples the joint optimization problem mathematically, enabling tractable learning of device selection policies while delegating resource allocation to convex optimization tools. To improve adaptability, T-ELLM integrates a sample-efficient, model-based virtual learning environment that captures the relationship between device selection and learning performance, facilitating subsequent group relative policy optimization. This concerted approach reduces reliance on real-world interactions, minimizing communication overhead while maintaining high-fidelity decision-making. Theoretical analysis proves that the discrepancy between virtual and real environments is bounded, ensuring the advantage function learned in the virtual environment maintains a provably small deviation from real-world conditions. Experimental results demonstrate that T-ELLM outperforms benchmark methods in energy efficiency and exhibits robust adaptability to environmental changes.
Large Language Model (LLM) Agents have demonstrated remarkable capabilities in task automation and intelligent decision-making, driving the widespread adoption of agent development frameworks such as LangChain and AutoGen. However, these frameworks predominantly serve developers with extensive technical expertise - a significant limitation considering that only 0.03 % of the global population possesses the necessary programming skills. This stark accessibility gap raises a fundamental question: Can we enable everyone, regardless of technical background, to build their own LLM agents using natural language alone? To address this challenge, we introduce AutoAgent-a Fully-Automated and highly Self-Developing framework that enables users to create and deploy LLM agents through Natural Language Alone. Operating as an autonomous Agent Operating System, AutoAgent comprises four key components: i) Agentic System Utilities, ii) LLM-powered Actionable Engine, iii) Self-Managing File System, and iv) Self-Play Agent Customization module. This lightweight yet powerful system enables efficient and dynamic creation and modification of tools, agents, and workflows without coding requirements or manual intervention. Beyond its code-free agent development capabilities, AutoAgent also serves as a versatile multi-agent system for General AI Assistants. Comprehensive evaluations on the GAIA benchmark demonstrate AutoAgent's effectiveness in generalist multi-agent tasks, surpassing existing state-of-the-art methods. Furthermore, AutoAgent's Retrieval-Augmented Generation (RAG)-related capabilities have shown consistently superior performance compared to many alternative LLM-based solutions.
Query routing, the task to route user queries to different large language model (LLM) endpoints, can be considered as a text classification problem. However, out-of-distribution queries must be handled properly, as those could be questions about unrelated domains, queries in other languages, or even contain unsafe text. Here, we thus study a \emph{guarded} query routing problem, for which we first introduce the Guarded Query Routing Benchmark (GQR-Bench), which covers three exemplary target domains (law, finance, and healthcare), and seven datasets to test robustness against out-of-distribution queries. We then use GQR-Bench to contrast the effectiveness and efficiency of LLM-based routing mechanisms (GPT-4o-mini, Llama-3.2-3B, and Llama-3.1-8B), standard LLM-based guardrail approaches (LlamaGuard and NVIDIA NeMo Guardrails), continuous bag-of-words classifiers (WideMLP, fastText), and traditional machine learning models (SVM, XGBoost). Our results show that WideMLP, enhanced with out-of-domain detection capabilities, yields the best trade-off between accuracy (88\%) and speed (<4ms). The embedding-based fastText excels at speed (<1ms) with acceptable accuracy (80\%), whereas LLMs yield the highest accuracy (91\%) but are comparatively slow (62ms for local Llama-3.1:8B and 669ms for remote GPT-4o-mini calls). Our findings challenge the automatic reliance on LLMs for (guarded) query routing and provide concrete recommendations for practical applications. GQR-Bench will be released as a Python package -- \texttt{gqr}.
Large language model (LLM) agents have shown promising performance in generating code for solving complex data science problems. Recent studies primarily focus on enhancing in-context learning through improved search, sampling, and planning techniques, while overlooking the importance of the order in which problems are tackled during inference. In this work, we develop a novel inference-time optimization framework, referred to as DSMentor, which leverages curriculum learning -- a strategy that introduces simpler task first and progressively moves to more complex ones as the learner improves -- to enhance LLM agent performance in challenging data science tasks. Our mentor-guided framework organizes data science tasks in order of increasing difficulty and incorporates a growing long-term memory to retain prior experiences, guiding the agent's learning progression and enabling more effective utilization of accumulated knowledge. We evaluate DSMentor through extensive experiments on DSEval and QRData benchmarks. Experiments show that DSMentor using Claude-3.5-Sonnet improves the pass rate by up to 5.2% on DSEval and QRData compared to baseline agents. Furthermore, DSMentor demonstrates stronger causal reasoning ability, improving the pass rate by 8.8% on the causality problems compared to GPT-4 using Program-of-Thoughts prompts. Our work underscores the importance of developing effective strategies for accumulating and utilizing knowledge during inference, mirroring the human learning process and opening new avenues for improving LLM performance through curriculum-based inference optimization.
Evaluating ecological time series is critical for benchmarking model performance in many important applications, including predicting greenhouse gas fluxes, capturing carbon-nitrogen dynamics, and monitoring hydrological cycles. Traditional numerical metrics (e.g., R-squared, root mean square error) have been widely used to quantify the similarity between modeled and observed ecosystem variables, but they often fail to capture domain-specific temporal patterns critical to ecological processes. As a result, these methods are often accompanied by expert visual inspection, which requires substantial human labor and limits the applicability to large-scale evaluation. To address these challenges, we propose a novel framework that integrates metric learning with large language model (LLM)-based natural language policy extraction to develop interpretable evaluation criteria. The proposed method processes pairwise annotations and implements a policy optimization mechanism to generate and combine different assessment metrics. The results obtained on multiple datasets for evaluating the predictions of crop gross primary production and carbon dioxide flux have confirmed the effectiveness of the proposed method in capturing target assessment preferences, including both synthetically generated and expert-annotated model comparisons. The proposed framework bridges the gap between numerical metrics and expert knowledge while providing interpretable evaluation policies that accommodate the diverse needs of different ecosystem modeling studies.
Long chain-of-thought (Long-CoT) reasoning improves accuracy in LLMs, yet its verbose, self-reflective style often hinders effective distillation into small language models (SLMs). We revisit Long-CoT compression through the lens of capability alignment and ask: Can pruning improve reasoning? We propose Prune-on-Logic, a structure-aware framework that transforms Long-CoT into logic graphs and selectively prunes low-utility reasoning steps under self-verification constraints. Through systematic analysis across three pruning strategies -- targeting entire chains, core reasoning, and verification -- we find that pruning verification steps yields consistent accuracy gains while reducing inference cost, outperforming token-level baselines and uncompressed fine-tuning. In contrast, pruning reasoning or all-chain steps degrades performance, revealing that small models benefit not from shorter CoTs, but from semantically leaner ones. Our findings highlight pruning as a structural optimization strategy for aligning CoT reasoning with SLM capacity.
Research on holistic Automated Essay Scoring (AES) is long-dated; yet, there is a notable lack of attention for assessing essays according to individual traits. In this work, we propose TRATES, a novel trait-specific and rubric-based cross-prompt AES framework that is generic yet specific to the underlying trait. The framework leverages a Large Language Model (LLM) that utilizes the trait grading rubrics to generate trait-specific features (represented by assessment questions), then assesses those features given an essay. The trait-specific features are eventually combined with generic writing-quality and prompt-specific features to train a simple classical regression model that predicts trait scores of essays from an unseen prompt. Experiments show that TRATES achieves a new state-of-the-art performance across all traits on a widely-used dataset, with the generated LLM-based features being the most significant.
In this paper, we present FuxiMT, a novel Chinese-centric multilingual machine translation model powered by a sparsified large language model (LLM). We adopt a two-stage strategy to train FuxiMT. We first pre-train the model on a massive Chinese corpus and then conduct multilingual fine-tuning on a large parallel dataset encompassing 65 languages. FuxiMT incorporates Mixture-of-Experts (MoEs) and employs a curriculum learning strategy for robust performance across various resource levels. Experimental results demonstrate that FuxiMT significantly outperforms strong baselines, including state-of-the-art LLMs and machine translation models, particularly under low-resource scenarios. Furthermore, FuxiMT exhibits remarkable zero-shot translation capabilities for unseen language pairs, indicating its potential to bridge communication gaps where parallel data are scarce or unavailable.
Large language models (LLMs) exhibit strong capabilities as decision-making agents by interleaving reasoning and actions, as seen in ReAct-style frameworks. Yet, their practical deployment is constrained by high inference costs and large model sizes. We propose Structured Agent Distillation, a framework that compresses large LLM-based agents into smaller student models while preserving both reasoning fidelity and action consistency. Unlike standard token-level distillation, our method segments trajectories into {[REASON]} and {[ACT]} spans, applying segment-specific losses to align each component with the teacher's behavior. This structure-aware supervision enables compact agents to better replicate the teacher's decision process. Experiments on ALFWorld, HotPotQA-ReAct, and WebShop show that our approach consistently outperforms token-level and imitation learning baselines, achieving significant compression with minimal performance drop. Scaling and ablation results further highlight the importance of span-level alignment for efficient and deployable agents.
As Large Language Models (LLMs) are frequently updated, LoRA weights trained on earlier versions quickly become obsolete. The conventional practice of retraining LoRA weights from scratch on the latest model is costly, time-consuming, and environmentally detrimental, particularly as the diversity of LLMs and downstream tasks expands. This motivates a critical question: "How can we efficiently leverage existing LoRA weights to adapt to newer model versions?" To address this, we propose LoRASuite, a modular approach tailored specifically to various types of LLM updates. First, we compute a transfer matrix utilizing known parameters from both old and new LLMs. Next, we allocate corresponding layers and attention heads based on centered kernel alignment and cosine similarity metrics, respectively. A subsequent small-scale, skillful fine-tuning step ensures numerical stability. Experimental evaluations demonstrate that LoRASuite consistently surpasses small-scale vanilla LoRA methods. Notably, on backbone LLMs such as MiniCPM and Qwen, LoRASuite even exceeds the performance of full-scale LoRA retraining, with average improvements of +1.4 and +6.6 points on math tasks, respectively. Additionally, LoRASuite significantly reduces memory consumption by 5.5 GB and computational time by 78.23%.
Knowledge tracing (KT) aims to predict learners' future performance based on historical learning interactions. However, existing KT models predominantly focus on data from a single course, limiting their ability to capture a comprehensive understanding of learners' knowledge states. In this paper, we propose TransKT, a contrastive cross-course knowledge tracing method that leverages concept graph guided knowledge transfer to model the relationships between learning behaviors across different courses, thereby enhancing knowledge state estimation. Specifically, TransKT constructs a cross-course concept graph by leveraging zero-shot Large Language Model (LLM) prompts to establish implicit links between related concepts across different courses. This graph serves as the foundation for knowledge transfer, enabling the model to integrate and enhance the semantic features of learners' interactions across courses. Furthermore, TransKT includes an LLM-to-LM pipeline for incorporating summarized semantic features, which significantly improves the performance of Graph Convolutional Networks (GCNs) used for knowledge transfer. Additionally, TransKT employs a contrastive objective that aligns single-course and cross-course knowledge states, thereby refining the model's ability to provide a more robust and accurate representation of learners' overall knowledge states.
We introduce UniGen, a unified multimodal large language model (MLLM) capable of image understanding and generation. We study the full training pipeline of UniGen from a data-centric perspective, including multi-stage pre-training, supervised fine-tuning, and direct preference optimization. More importantly, we propose a new Chain-of-Thought Verification (CoT-V) strategy for test-time scaling, which significantly boosts UniGen's image generation quality using a simple Best-of-N test-time strategy. Specifically, CoT-V enables UniGen to act as both image generator and verifier at test time, assessing the semantic alignment between a text prompt and its generated image in a step-by-step CoT manner. Trained entirely on open-source datasets across all stages, UniGen achieves state-of-the-art performance on a range of image understanding and generation benchmarks, with a final score of 0.78 on GenEval and 85.19 on DPG-Bench. Through extensive ablation studies, our work provides actionable insights and addresses key challenges in the full life cycle of building unified MLLMs, contributing meaningful directions to the future research.
While large language model (LLM)-based chatbots have demonstrated strong capabilities in generating coherent and contextually relevant responses, they often struggle with understanding when to speak, particularly in delivering brief, timely reactions during ongoing conversations. This limitation arises largely from their reliance on text input, lacking the rich contextual cues in real-world human dialogue. In this work, we focus on real-time prediction of response types, with an emphasis on short, reactive utterances that depend on subtle, multimodal signals across vision, audio, and text. To support this, we introduce a new multimodal dataset constructed from real-world conversational videos, containing temporally aligned visual, auditory, and textual streams. This dataset enables fine-grained modeling of response timing in dyadic interactions. Building on this dataset, we propose MM-When2Speak, a multimodal LLM-based model that adaptively integrates visual, auditory, and textual context to predict when a response should occur, and what type of response is appropriate. Experiments show that MM-When2Speak significantly outperforms state-of-the-art unimodal and LLM-based baselines, achieving up to a 4x improvement in response timing accuracy over leading commercial LLMs. These results underscore the importance of multimodal inputs for producing timely, natural, and engaging conversational AI.
Large language model (LLM) based agents have shown great potential in following human instructions and automatically completing various tasks. To complete a task, the agent needs to decompose it into easily executed steps by planning. Existing studies mainly conduct the planning by inferring what steps should be executed next starting from the agent's initial state. However, this forward reasoning paradigm doesn't work well for complex tasks. We propose to study this issue in Minecraft, a virtual environment that simulates complex tasks based on real-world scenarios. We believe that the failure of forward reasoning is caused by the big perception gap between the agent's initial state and task goal. To this end, we leverage backward reasoning and make the planning starting from the terminal state, which can directly achieve the task goal in one step. Specifically, we design a BAckward Reasoning based agent (BAR). It is equipped with a recursive goal decomposition module, a state consistency maintaining module and a stage memory module to make robust, consistent, and efficient planning starting from the terminal state. Experimental results demonstrate the superiority of BAR over existing methods and the effectiveness of proposed modules.
Automated Essay Scoring (AES) is crucial for modern education, particularly with the increasing prevalence of multimodal assessments. However, traditional AES methods struggle with evaluation generalizability and multimodal perception, while even recent Multimodal Large Language Model (MLLM)-based approaches can produce hallucinated justifications and scores misaligned with human judgment. To address the limitations, we introduce CAFES, the first collaborative multi-agent framework specifically designed for AES. It orchestrates three specialized agents: an Initial Scorer for rapid, trait-specific evaluations; a Feedback Pool Manager to aggregate detailed, evidence-grounded strengths; and a Reflective Scorer that iteratively refines scores based on this feedback to enhance human alignment. Extensive experiments, using state-of-the-art MLLMs, achieve an average relative improvement of 21% in Quadratic Weighted Kappa (QWK) against ground truth, especially for grammatical and lexical diversity. Our proposed CAFES framework paves the way for an intelligent multimodal AES system. The code will be available upon acceptance.
Embodied Question Answering (EQA) requires agents to autonomously explore and understand the environment to answer context-dependent questions. Existing frameworks typically center around the planner, which guides the stopping module, memory module, and answering module for reasoning. In this paper, we propose a memory-centric EQA framework named MemoryEQA. Unlike planner-centric EQA models where the memory module cannot fully interact with other modules, MemoryEQA flexible feeds memory information into all modules, thereby enhancing efficiency and accuracy in handling complex tasks, such as those involving multiple targets across different regions. Specifically, we establish a multi-modal hierarchical memory mechanism, which is divided into global memory that stores language-enhanced scene maps, and local memory that retains historical observations and state information. When performing EQA tasks, the multi-modal large language model is leveraged to convert memory information into the required input formats for injection into different modules. To evaluate EQA models' memory capabilities, we constructed the MT-HM3D dataset based on HM3D, comprising 1,587 question-answer pairs involving multiple targets across various regions, which requires agents to maintain memory of exploration-acquired target information. Experimental results on HM-EQA, MT-HM3D, and OpenEQA demonstrate the effectiveness of our framework, where a 19.8% performance gain on MT-HM3D compared to baseline model further underscores memory capability's pivotal role in resolving complex tasks.
We introduce Llama-Krikri-8B, a cutting-edge Large Language Model tailored for the Greek language, built on Meta's Llama 3.1-8B. Llama-Krikri-8B has been extensively trained on high-quality Greek data to ensure superior adaptation to linguistic nuances. With 8 billion parameters, it offers advanced capabilities while maintaining efficient computational performance. Llama-Krikri-8B supports both Modern Greek and English, and is also equipped to handle polytonic text and Ancient Greek. The chat version of Llama-Krikri-8B features a multi-stage post-training pipeline, utilizing both human and synthetic instruction and preference data, by applying techniques such as MAGPIE. In addition, for evaluation, we propose three novel public benchmarks for Greek. Our evaluation on existing as well as the proposed benchmarks shows notable improvements over comparable Greek and multilingual LLMs in both natural language understanding and generation as well as code generation.
Scientific retrieval is essential for advancing academic discovery. Within this process, document reranking plays a critical role by refining first-stage retrieval results. However, large language model (LLM) listwise reranking faces unique challenges in the scientific domain. First-stage retrieval is often suboptimal in the scientific domain, so relevant documents are ranked lower. Moreover, conventional listwise reranking uses the full text of candidate documents in the context window, limiting the number of candidates that can be considered. As a result, many relevant documents are excluded before reranking, which constrains overall retrieval performance. To address these challenges, we explore compact document representations based on semantic features such as categories, sections, and keywords, and propose a training-free, model-agnostic reranking framework for scientific retrieval called CoRank. The framework involves three stages: (i) offline extraction of document-level features, (ii) coarse reranking using these compact representations, and (iii) fine-grained reranking on full texts of the top candidates from stage (ii). This hybrid design provides a high-level abstraction of document semantics, expands candidate coverage, and retains critical details required for precise ranking. Experiments on LitSearch and CSFCube show that CoRank significantly improves reranking performance across different LLM backbones, increasing nDCG@10 from 32.0 to 39.7. Overall, these results highlight the value of information extraction for reranking in scientific retrieval.