Large language models (LLMs) integrated with retrieval-augmented generation (RAG) systems improve accuracy by leveraging external knowledge sources. However, recent research has revealed RAG's susceptibility to poisoning attacks, where the attacker injects poisoned texts into the knowledge database, leading to attacker-desired responses. Existing defenses, which predominantly focus on inference-time mitigation, have proven insufficient against sophisticated attacks. In this paper, we introduce RAGForensics, the first traceback system for RAG, designed to identify poisoned texts within the knowledge database that are responsible for the attacks. RAGForensics operates iteratively, first retrieving a subset of texts from the database and then utilizing a specially crafted prompt to guide an LLM in detecting potential poisoning texts. Empirical evaluations across multiple datasets demonstrate the effectiveness of RAGForensics against state-of-the-art poisoning attacks. This work pioneers the traceback of poisoned texts in RAG systems, providing a practical and promising defense mechanism to enhance their security. Our code is available at: https://github.com/zhangbl6618/RAG-Responsibility-Attribution

With the rapid expansion of content on social media platforms, analyzing and comprehending online discourse has become increasingly complex. This paper introduces LLMTaxo, a novel framework leveraging large language models for the automated construction of taxonomies of factual claims from social media by generating topics at multiple levels of granularity. The resulting hierarchical structure significantly reduces redundancy and improves information accessibility. We also propose dedicated taxonomy evaluation metrics to enable comprehensive assessment. Evaluations conducted on three diverse datasets demonstrate LLMTaxo's effectiveness in producing clear, coherent, and comprehensive taxonomies. Among the evaluated models, GPT-4o mini consistently outperforms others across most metrics. The framework's flexibility and low reliance on manual intervention underscore its potential for broad applicability.

Cross-lingual context retrieval (extracting contextual information in one language based on requests in another) is a fundamental aspect of cross-lingual alignment, but the performance and mechanism of it for large language models (LLMs) remains unclear. In this paper, we evaluate the cross-lingual context retrieval of over 40 LLMs across 12 languages, using cross-lingual machine reading comprehension (xMRC) as a representative scenario. Our results show that post-trained open LLMs show strong cross-lingual context retrieval ability, comparable to closed-source LLMs such as GPT-4o, and their estimated oracle performances greatly improve after post-training. Our mechanism analysis shows that the cross-lingual context retrieval process can be divided into two main phases: question encoding and answer retrieval, which are formed in pre-training and post-training respectively. The phasing stability correlates with xMRC performance, and the xMRC bottleneck lies at the last model layers in the second phase, where the effect of post-training can be evidently observed. Our results also indicate that larger-scale pretraining cannot improve the xMRC performance. Instead, larger LLMs need further multilingual post-training to fully unlock their cross-lingual context retrieval potential.

Large reasoning models (LRMs) have recently demonstrated impressive capabilities in complex reasoning tasks by leveraging increased test-time computation and exhibiting behaviors reminiscent of human-like self-reflection. While LRMs show a clear capacity for valuable self-reflection, how this ability interacts with other model behaviors remains underexplored. We investigate this connection by analyzing verbalized confidence, how models articulate their certainty, as a lens into the nature of self-reflection in LRMs. We find that supervised fine-tuning on reasoning traces (i.e., distillation) and reinforcement learning can improve verbalized calibration in reasoning-intensive settings in a progressive, laddered fashion. However, our results also indicate that reasoning models may possess a diminished awareness of their own knowledge boundaries, as evidenced by significantly lower "I don't know" response rates on factuality benchmarks. Moreover, we examine the relationship between verbalized confidence and reasoning chains, finding that models tend to express higher confidence when providing shorter or less elaborate reasoning. Our findings highlight how reasoning-oriented training can enhance performance in reasoning-centric tasks while potentially incurring a "reasoning tax," a cost reflected in the model's reduced ability to accurately recognize the limits of its own knowledge in small-scale models. More broadly, our work showcases how this erosion of knowledge boundaries can compromise model faithfulness, as models grow more confident without a commensurate understanding of when they should abstain.

Preference alignment methods are increasingly critical for steering large language models (LLMs) to generate outputs consistent with human values. While recent approaches often rely on synthetic data generated by LLMs for scalability and cost-efficiency reasons, this reliance can introduce distribution shifts that undermine the nuanced representation of human preferences needed for desirable outputs. In this paper, we propose a novel distribution-aware optimization framework that improves preference alignment despite such shifts. Our approach first leverages well-learned classifiers to assign a calibration value to each training sample, quantifying its alignment with the target human-preferred distribution. These values are then incorporated into a robust optimization objective that minimizes the worst-case loss over regions of the data space most relevant to human preferences. By explicitly focusing optimization on the target distribution, our approach mitigates the impact of distributional mismatch and improves the generation of responses that better reflect intended values.

The widespread adoption of language models (LMs) has caused a huge surge in demand for GPUs. Training large LMs requires tens of thousands of GPUs and housing them in the same datacenter (DC) is a challenge due to many constraints including availability of peak power. We focus on training such models across multiple DCs connected via the Wide-Area-Network (WAN). We built Atlas that speeds up the training time using novel workload-aware temporal bandwidth sharing and other design choices. While Atlas improves the training time, it does not completely eliminate the bubbles (idle GPU cycles). We built BubbleTea that runs prefill-as-a-service (part of LM inference) during the bubbles thus improving the GPU utilization without any impact on training. Compared to state-of-the-art designs, Atlas and BubbleTea together achieve up to 17x faster training, and up to 94% GPU utilization. The code will be open-sourced.

Vision-Language Models (VLMs) achieve strong results on multimodal tasks such as visual question answering, yet they can still fail even when the correct visual evidence is present. In this work, we systematically investigate whether these failures arise from not perceiving the evidence or from not leveraging it effectively. By examining layer-wise attention dynamics, we find that shallow layers focus primarily on text, while deeper layers sparsely but reliably attend to localized evidence regions. Surprisingly, VLMs often perceive the visual evidence when outputting incorrect answers, a phenomenon we term ``seeing but not believing'' that widely exists in major VLM families. Building on this, we introduce an inference-time intervention that highlights deep-layer evidence regions through selective attention-based masking. It requires no training and consistently improves accuracy across multiple families, including LLaVA, Qwen, Gemma, and InternVL. These results show that VLMs encode reliable evidence internally but under-utilize it, making such signals explicit can bridge the gap between perception and reasoning, advancing the diagnostic understanding and reliability of VLMs.

Medical Large language models achieve strong scores on standard benchmarks; however, the transfer of those results to safe and reliable performance in clinical workflows remains a challenge. This survey reframes evaluation through a levels-of-autonomy lens (L0-L3), spanning informational tools, information transformation and aggregation, decision support, and supervised agents. We align existing benchmarks and metrics with the actions permitted at each level and their associated risks, making the evaluation targets explicit. This motivates a level-conditioned blueprint for selecting metrics, assembling evidence, and reporting claims, alongside directions that link evaluation to oversight. By centering autonomy, the survey moves the field beyond score-based claims toward credible, risk-aware evidence for real clinical use.

Abstract--Named Entity Recognition (NER) is a critical task that requires substantial annotated data, making it challenging in low-resource scenarios where label acquisition is expensive. While zero-shot and instruction-tuned approaches have made progress, they often fail to generalize to domain-specific entities and do not effectively utilize limited available data. We present a lightweight few-shot NER framework that addresses these challenges through two key innovations: (1) a new instruction tuning template with a simplified output format that combines principles from prior IT approaches to leverage the large context window of recent state-of-the-art LLMs; (2) introducing a strategic data augmentation technique that preserves entity information while paraphrasing the surrounding context, thereby expanding our training data without compromising semantic relationships. Experiments on benchmark datasets show that our method achieves performance comparable to state-of-the-art models on few-shot and zero-shot tasks, with our few-shot approach attaining an average F1 score of 80.1 on the CrossNER datasets. Models trained with our paraphrasing approach show consistent improvements in F1 scores of up to 17 points over baseline versions, offering a promising solution for groups with limited NER training data and compute power . Index T erms--Named Entity Recognition (NER), Few-Shot Learning, Large Language Models (LLMs), Instruction T uning, Data Augmentation. Named Entity Recognition (NER) is a foundational task in Natural Language Processing (NLP), enabling applications like information extraction, question answering, and event detection [1]. Traditional NER systems rely on supervised learning, requiring extensive annotated data for specific domains and predefined entity types. This dependency on large, labelled datasets limits their adaptability to new domains and entity categories.

Large Language Models have achieved strong performance on reasoning tasks, solving competition-level coding and math problems. However, their scalability is limited by human-labeled datasets and the lack of large-scale, challenging coding problem training data. Existing competitive coding datasets contain only thousands to tens of thousands of problems. Previous synthetic data generation methods rely on either augmenting existing instruction datasets or selecting challenging problems from human-labeled data. In this paper, we propose QueST, a novel framework which combines difficulty-aware graph sampling and difficulty-aware rejection fine-tuning that directly optimizes specialized generators to create challenging coding problems. Our trained generators demonstrate superior capability compared to even GPT-4o at creating challenging problems that benefit downstream performance. We leverage QueST to generate large-scale synthetic coding problems, which we then use to distill from strong teacher models with long chain-of-thought or to conduct reinforcement learning for smaller models, proving effective in both scenarios. Our distillation experiments demonstrate significant performance gains. Specifically, after fine-tuning Qwen3-8B-base on 100K difficult problems generated by QueST, we surpass the performance of the original Qwen3-8B on LiveCodeBench. With an additional 112K examples (i.e., 28K human-written problems paired with multiple synthetic solutions), our 8B model matches the performance of the much larger DeepSeek-R1-671B. These findings indicate that generating complex problems via QueST offers an effective and scalable approach to advancing the frontiers of competitive coding and reasoning for large language models.