Paper · Needs Review
User Feedback Provides a Unique Signal that LLMs Can not DetectUser feedback helps models repair targeted faults, but LLM judges often miss those repairs. Agent evals should retain human feedback and avoid treating model preference as ground truth.FEED7 SUMMARY
Paper · Needs Review
The Implications of Linguistic Illegibility for LLM SecurityLanguage-based monitoring cannot fully reveal an agent’s internal computation, the paper argues. Secure agent runtimes still need isolation and data-flow controls independent of self-reporting.FEED7 SUMMARY
infra#sandboxing
Paper · Needs Review
Post-Training Language Models for Gold-Medal Performance in Coding CompetitionsA coding-specialized model paired post-training with an iterative generate-evaluate-refine loop to exceed the top IOI 2026 human score. The reusable idea is feedback-driven test-time search.FEED7 SUMMARY
model#reasoning
Paper · Needs Review
UE5M3 FP4 Block Scaling for Stable Language Model PretrainingA UE5M3 block-scaling recipe trained an 8B model in FP4 without Hadamard transforms or BF16 final layers, while reporting better losses and downstream estimates than the compared recipe.FEED7 SUMMARY
model#model-selection
Paper · Needs Review
Cliff: Learning Process Rewards from the First MistakeCliff trains on the location of a rollout’s first reasoning error, rewarding the valid prefix and penalizing the suffix. It offers finer supervision without a specialized reward model.FEED7 SUMMARY
model#reasoning
Paper · Needs Review
Large Language Models (LLMs) for Telecom Root Cause Analysis (RCA): A Structured Reasoning Framework for Evidence-Grounded DiagnosisA telecom diagnosis pipeline structures telemetry before forcing a decision path and evidence-grounded explanation. The pattern is relevant to agents reasoning over heterogeneous operational data.FEED7 SUMMARY
Paper · Needs Review
Dutch Books for Language ModelsA label-free Dutch-book test finds internally inconsistent probabilities from language models, especially when prompts add logical complexity or irrelevant context.FEED7 SUMMARY
Paper · Needs Review
SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety AlignmentSafeEvolve turns completed agent trajectories into reversible harness updates and policy training, linking runtime safeguards with learned multi-step behavior.FEED7 SUMMARY
agent#harness-engineering
Paper · Needs Review
Beyond Scores: Understanding LLM-as-a-Judge Mechanisms in Summarization EvaluationMechanistic tests suggest two open LLM judges route local errors early and form ratings late, giving eval builders concrete layers and components to probe for judge failures.FEED7 SUMMARY
Paper · Needs Review
CordisBench: Can Language Models Reason About Component Lifecycles in Dynamic Agent Harnesses?CordisBench finds that models lose reliability as harness lifecycle interactions grow, while deterministic execution semantics can replace costly reasoning on controlled cases.FEED7 SUMMARY
benchmark#harness-engineering
Paper · Needs Review
The Rise of Verbal Reinforcement LearningThis survey separates verbal feedback by when it acts and what it changes, helping agent builders distinguish task specification, test-time guidance, and parameter updates.FEED7 SUMMARY
agent#prompting
Paper · Needs Review
StudentSim: Training LLM-based Student SimulatorsStudentSim turns sparse user histories into individualized simulators that model both current behavior and response to guidance, a pattern for testing adaptive agents before live deployment.FEED7 SUMMARY