GraphReview: Scientific Paper Evaluation via LLM-Based Graph Message Passing
GraphReview proposes a graph-based LLM framework that models scientific paper evaluation as review-signal message passing over a semantic paper graph, capturing both intrinsic quality and relational context (synchronic and diachronic links). LLMs estimate node-level quality priors and generate edge-level comparative evidence via pairwise comparisons, while Personalized PageRank integrates signals for ranking, decision prediction, and review generation. The system uses reward-induced maximum likelihood objectives to train LLM backbones and achieves average improvements of 29.7% over the strongest baseline on decision and ranking metrics, including 23.7% accuracy gain and 57.6% Spearman's ρ gain.
Related guides (2)
Related events (8)
ReproRepo: Scalable LLM agent framework for reproducibility auditing using GitHub issues
ReproRepo is a new framework for evaluating LLM agents on reproducibility auditing of ML research, using naturally occurring GitHub issues as supervision signals rather than costly manual curation. The framework is instantiated on 1,149 recent ML papers from major conferences and benchmarks four frontier model-agent configurations. The best-performing agent (Codex with GPT-5.5) surfaces at least one semantically related human-reported reproduction blocker for ~90% of papers, though exact localization of issues remains a weakness. The work provides a reusable, scalable evaluation harness for this underexplored agentic task.
SoundnessBench: Benchmarking LLMs as Evaluators of ML Research Proposal Viability
SoundnessBench is a new benchmark of 1,099 machine-learning research proposals derived from ICLR submissions, labeled with reviewer soundness scores, designed to test whether LLMs can reliably distinguish methodologically sound research ideas from unsound ones. Evaluated across 12 frontier LLMs, the benchmark reveals a pervasive optimism bias: models systematically rate low-soundness proposals as sound under standard prompting, with aggressive prompting shifting errors from false positives to false negatives rather than eliminating them. Controls for data contamination, surface features, and human audit quality suggest the bias is not attributable to a single confounder. The authors conclude that current LLMs are not yet reliable as standalone first-gate evaluators of scientific rigor, a critical bottleneck for autonomous AI research agents.
LLUMI: Fine-Tuning Open-Source LLMs for Mental Health Writing Assistance Using Reddit Community Feedback
LLUMI is a two-component system (a generation model and an improvement model) designed to provide mental health writing assistance using smaller open-source LLMs hosted in privacy-preserving, on-premise environments. The system leverages Reddit community endorsement signals (upvotes/downvotes) to construct preference pairs for SFT and DPO training, then further aligns outputs via human evaluation across readability, empathy, connection, actionability, and safety dimensions. Results show LLUMI achieves performance comparable to proprietary GPT-based models on linguistic and human evaluations, suggesting community-derived preference signals can substitute for expensive expert labeling in sensitive domains.
Mechanistic analysis of how LLMs encode essay quality in internal representations
Researchers systematically probe the hidden representations of eight LLMs across three essay datasets (ASAP++, CSEE, ENEM) to understand how automated essay scoring (AES) works internally. Using linear probing, dimensionality reduction, and neuron-level analysis, they find essay quality is encoded in a linearly accessible form that emerges progressively across layers and partially transfers across prompts. Individual 'essay scoring neurons' are identified whose activations correlate with scores and respond to targeted interventions, with longer essays relying more on deeper layers. The work contributes to mechanistic interpretability of LLM-based scoring systems.
PARL: Preference-Aware Rubric Learning for Personalized LLM Evaluation
This paper introduces PARL (Preference-Aware Rubric Learning), a framework that reframes personalized LLM evaluation as a learning problem rather than static judgment. PARL induces preference-aware evaluation rubrics from raw user interaction histories and uses a discriminative reinforcement learning objective to contrast user-authored responses against model outputs, capturing user-specific decision boundaries. Experiments on personalized text generation tasks show PARL produces high-fidelity rubrics that generalize across users and tasks, outperforming existing LLM-as-a-judge and automatic metric approaches.
Paper challenges LLM expert-level claims by measuring variance and error magnitude in code-based data analysis tasks
A new arXiv paper argues that standard LLM benchmarks overstate model capabilities by focusing on average performance on training-data-adjacent tasks while ignoring response variance and error magnitude. The authors introduce a novel benchmark requiring frontier LLMs to write code for data analysis tasks, comparing results against human expert submissions. Human experts outperformed the frontier LLM on average across multiple metrics and showed lower performance variability. The findings challenge the prevailing narrative that LLMs perform at human-expert level on knowledge economy tasks.
MetaSyn benchmark reveals critical screening bottleneck in LLM-based meta-analysis pipelines
Researchers introduce MetaSyn, a dataset of 442 expert-curated meta-analyses from Nature Portfolio journals, paired with a 140k-article PubMed retrieval corpus, PI/ECO criteria, verified positives, and hard negatives. Benchmarking twelve pipeline configurations — nine RAG variants and a protocol-driven agent — shows that despite 90.9% retrieval recall at K=200, no system recovers more than 52.7% of ground-truth included studies. The core failure is LLMs' inability to reliably distinguish eligible studies from topically similar but criteria-failing distractors. The paper argues that end-to-end scores obscure where pipelines break down and proposes stage-attributed metrics.
GraphPO: Graph-based Policy Optimization reduces redundancy in LLM reasoning RL
GraphPO is a new reinforcement learning framework that represents reasoning rollouts as directed acyclic graphs rather than independent chains or trees, merging semantically equivalent reasoning paths into equivalence classes to share suffixes and reduce redundant exploration. The approach assigns efficiency advantages to incoming edges and correctness advantages to outgoing edges, deriving process supervision from outcome rewards. Experiments on three LLMs across reasoning and agentic search benchmarks show consistent improvements over chain- and tree-based baselines under equal token or response budgets. The method also provides theoretical guarantees on reduced advantage-estimation variance.