Formalized Information Needs Improve Large-Language-Model Relevance Judgments

Cranfield-style retrieval evaluations with too few or too many relevant documents or with low inter-assessor agreement on relevance can reduce the reliability of observations. In evaluations with human assessors, information needs are often formalized as retrieval topics to avoid an excessive number of relevant documents while maintaining good agreement. However, emerging evaluation setups that use Large Language Models (LLMs) as relevance assessors often use only queries, potentially decreasing the reliability. To study whether LLM relevance assessors benefit from formalized information needs, we synthetically formalize information needs with LLMs into topics that follow the established structure from previous human relevance assessments (i.e., descriptions and narratives). We compare assessors using synthetically formalized topics against the LLM-default query-only assessor on Robust04 and the 2019/2020 editions of TREC Deep Learning. We find that assessors without formalization judge many more documents relevant and have a lower agreement, leading to reduced reliability in retrieval evaluations. Furthermore, we show that the formalized topics improve agreement between human and LLM relevance judgments, even when the topics are not highly similar to their human counterparts. Our findings indicate that LLM relevance assessors should use formalized information needs, as is standard for human assessment, and synthetically formalize topics when no human formalization exists to improve evaluation reliability.

Self-Preference Bias in Rubric-Based Evaluation of Large Language Models

LLM-as-a-judge has become the de facto approach for evaluating LLM outputs. However, judges are known to exhibit self-preference bias (SPB): they tend to favor outputs produced by themselves or by models from their own family. This skews evaluations and, thus, hinders model development, especially in settings of recursive self-improvement. We present the first study of SPB in rubric-based evaluation, an increasingly popular benchmarking paradigm where judges issue binary verdicts on individual evaluation criteria, instead of assigning holistic scores or rankings. Using IFEval, a benchmark with programmatically verifiable rubrics, we show that SPB persists even when evaluation criteria are entirely objective: among rubrics where generators fail, judges can be up to 50\% more likely to incorrectly mark them as satisfied when the output is their own. We also find that, similarly to other evaluation paradigms, ensembling multiple judges helps mitigate SPB, but without fully eliminating it. On HealthBench, a medical chat benchmark with subjective rubrics, we observe that SPB skews model scores by up to 10 points, a potentially decisive margin when ranking frontier models. We analyze the factors that drive SPB in this setting, finding that negative rubrics, extreme rubric lengths, and subjective topics like emergency referrals are particularly susceptible.

GAM: Hierarchical Graph-based Agentic Memory for LLM Agents

To sustain coherent long-term interactions, Large Language Model (LLM) agents must navigate the tension between acquiring new information and retaining prior knowledge. Current unified stream-based memory systems facilitate context updates but remain vulnerable to interference from transient noise. Conversely, discrete structured memory architectures provide robust knowledge retention but often struggle to adapt to evolving narratives. To address this, we propose GAM, a hierarchical Graph-based Agentic Memory framework that explicitly decouples memory encoding from consolidation to effectively resolve the conflict between rapid context perception and stable knowledge retention. By isolating ongoing dialogue in an event progression graph and integrating it into a topic associative network only upon semantic shifts, our approach minimizes interference while preserving long-term consistency. Additionally, we introduce a graph-guided, multi-factor retrieval strategy to enhance context precision. Experiments on LoCoMo and LongDialQA indicate that our method consistently outperforms state-of-the-art baselines in both reasoning accuracy and efficiency.

Talk2AI: A Longitudinal Dataset of Human--AI Persuasive Conversations

Talk2AI is a large-scale longitudinal dataset of 3,080 conversations (totaling 30,800 turns) between human participants and Large Language Models (LLMs), designed to support research on persuasion, opinion change, and human-AI interaction. The corpus was collected from 770 profiled Italian adults across four weekly sessions in Spring 2025, using a within-subject design in which each participant conversed with a single model (GPT-4o, Claude Sonnet 3.7, DeepSeek-chat V3, or Mistral Large) on three socially relevant topics: climate change, math anxiety, and health misinformation. Each conversation is linked to rich contextual data, including sociodemographic characteristics and psychometric profiles. After each session, participants reported on opinion change, conviction stability, perceived humanness of the AI, and behavioral intentions, enabling fine-grained longitudinal analysis of how AI-mediated dialogue shapes beliefs and attitudes over time.

Case-Grounded Evidence Verification: A Framework for Constructing Evidence-Sensitive Supervision

Evidence-grounded reasoning requires more than attaching retrieved text to a prediction: a model should make decisions that depend on whether the provided evidence supports the target claim. In practice, this often fails because supervision is weak, evidence is only loosely tied to the claim, and evaluation does not test evidence dependence directly. We introduce case-grounded evidence verification, a general framework in which a model receives a local case context, external evidence, and a structured claim, and must decide whether the evidence supports the claim for that case. Our key contribution is a supervision construction procedure that generates explicit support examples together with semantically controlled non-support examples, including counterfactual wrong-state and topic-related negatives, without manual evidence annotation. We instantiate the framework in radiology and train a standard verifier on the resulting support task. The learned verifier substantially outperforms both case-only and evidence-only baselines, remains strong under correct evidence, and collapses when evidence is removed or swapped, indicating genuine evidence dependence. This behavior transfers across unseen evidence articles and an external case distribution, though performance degrades under evidence-source shift and remains sensitive to backbone choice. Overall, the results suggest that a major bottleneck in evidence grounding is not only model capacity, but the lack of supervision that encodes the causal role of evidence.

Multilingual Embedding Probes Fail to Generalize Across Learner Corpora

Do multilingual embedding models encode a language-general representation of proficiency? We investigate this by training linear and non-linear probes on hidden-state activations from Qwen3-Embedding (0.6B, 4B, 8B) to predict CEFR proficiency levels from learner texts across nine corpora and seven languages. We compare five probing architectures against a baseline trained on surface-level text features. Under in-distribution evaluation, probes achieve strong performance ($QWK\approx0.7$), substantially outperforming the surface baseline, with middle layers consistently yielding the best predictions. However, in cross-corpus evaluation performance collapses across all probe types and model sizes. Residual analysis reveals that out-of-distribution probes converge towards predicting uniformly distributed labels, indicating that the learned mappings capture corpus-specific distributional properties (topic, language, task type, rating methodology) rather than an abstract, transferable proficiency dimension. These results suggest that current multilingual embeddings do not straightforwardly encode language-general proficiency, with implications for representation-based approaches to proficiency-adaptive language technology.

How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models

This paper identifies a recurring sparse routing mechanism in alignment-trained language models: a gate attention head reads detected content and triggers downstream amplifier heads that boost the signal toward refusal. Using political censorship and safety refusal as natural experiments, the mechanism is traced across 9 models from 6 labs, all validated on corpora of 120 prompt pairs. The gate head passes necessity and sufficiency interchange tests (p < 0.001, permutation null), and core amplifier heads are stable under bootstrap resampling (Jaccard 0.92-1.0). Three same-generation scaling pairs show that routing distributes at scale (ablation up to 17x weaker) while remaining detectable by interchange. Modulating the detection-layer signal continuously controls policy strength from hard refusal through steering to factual compliance, with routing thresholds that vary by topic. The circuit also reveals a structural separation between intent recognition and policy routing: under cipher encoding, the gate head's interchange necessity collapses 70-99% across three models (n=120), and the model responds with puzzle-solving rather than refusal. The routing mechanism never fires, even though probe scores at deeper layers indicate the model begins to represent the harmful content. This asymmetry is consistent with different robustness properties of pretraining and post-training: broad semantic understanding versus narrower policy binding that generalizes less well under input transformation.

Beyond Relevance: Utility-Centric Retrieval in the LLM Era

Information retrieval systems have traditionally optimized for topical relevance-the degree to which retrieved documents match a query. However, relevance only approximates a deeper goal: utility, namely, whether retrieved information helps accomplish a user's underlying task. The emergence of retrieval-augmented generation (RAG) fundamentally changes this paradigm. Retrieved documents are no longer consumed directly by users but instead serve as evidence for large language models (LLMs) that produce answers. As a result, retrieval effectiveness must be evaluated by its contribution to generation quality rather than by relevance-based ranking metrics alone. This tutorial argues that retrieval objectives are evolving from relevance-centric optimization toward LLM-centric utility. We present a unified framework covering LLM-agnostic versus LLM-specific utility, context-independent versus context-dependent utility, and the connection with LLM information needs and agentic RAG. By synthesizing recent advances, the tutorial provides conceptual foundations and practical guidance for designing retrieval systems aligned with the requirements of LLM-based information access.

MisEdu-RAG: A Misconception-Aware Dual-Hypergraph RAG for Novice Math Teachers

Novice math teachers often encounter students' mistakes that are difficult to diagnose and remediate. Misconceptions are especially challenging because teachers must explain what went wrong and how to solve them. Although many existing large language model (LLM) platforms can assist in generating instructional feedback, these LLMs loosely connect pedagogical knowledge and student mistakes, which might make the guidance less actionable for teachers. To address this gap, we propose MisEdu-RAG, a dual-hypergraph-based retrieval-augmented generation (RAG) framework that organizes pedagogical knowledge as a concept hypergraph and real student mistake cases as an instance hypergraph. Given a query, MisEdu-RAG performs a two-stage retrieval to gather connected evidence from both layers and generates a response grounded in the retrieved cases and pedagogical principles. We evaluate on \textit{MisstepMath}, a dataset of math mistakes paired with teacher solutions, as a benchmark for misconception-aware retrieval and response generation across topics and error types. Evaluation results on \textit{MisstepMath} show that, compared with baseline models, MisEdu-RAG improves token-F1 by 10.95\% and yields up to 15.3\% higher five-dimension response quality, with the largest gains on \textit{Diversity} and \textit{Empowerment}. To verify its applicability in practical use, we further conduct a pilot study through a questionnaire survey of 221 teachers and interviews with 6 novices. The findings suggest that MisEdu-RAG provides diagnosis results and concrete teaching moves for high-demand misconception scenarios. Overall, MisEdu-RAG demonstrates strong potential for scalable teacher training and AI-assisted instruction for misconception handling. Our code is available on GitHub: https://github.com/GEMLab-HKU/MisEdu-RAG.

SocioEval: A Template-Based Framework for Evaluating Socioeconomic Status Bias in Foundation Models

As Large Language Models (LLMs) increasingly power decision-making systems across critical domains, understanding and mitigating their biases becomes essential for responsible AI deployment. Although bias assessment frameworks have proliferated for attributes such as race and gender, socioeconomic status bias remains significantly underexplored despite its widespread implications in the real world. We introduce SocioEval, a template-based framework for systematically evaluating socioeconomic bias in foundation models through decision-making tasks. Our hierarchical framework encompasses 8 themes and 18 topics, generating 240 prompts across 6 class-pair combinations. We evaluated 13 frontier LLMs on 3,120 responses using a rigorous three-stage annotation protocol, revealing substantial variation in bias rates (0.42\%-33.75\%). Our findings demonstrate that bias manifests differently across themes lifestyle judgments show 10$\times$ higher bias than education-related decisions and that deployment safeguards effectively prevent explicit discrimination but show brittleness to domain-specific stereotypes. SocioEval provides a scalable, extensible foundation for auditing class-based bias in language models.