ResearchAgentic Code Reasoning Semi-Formal Reasoning 88 Percent AccuracyarXiv·high signalStructured prompting for code reasoning without execution. 88% accuracy on patch verification, 93% on real-world patches. Verifiable certificates.SourceSource pagearXiv↳ Follow the threadSame source URL / Stack layerSemi-formal reasoning: force the agent to state premises and trace execution before it verifies a patcharXivStack layer / Threat patternLook-Ahead Bias Reframed as a Verifiable Non-Interference Property for Agentic TradingarXivStack layer / Threat patternPiSAs Benchmarks Cross-User Privacy Leaks in Shared Agentic SystemsarXivPolicy dependency / Stack layerTREK Rescues GRPO on Hard Prompts Outside the Student's On-Policy SupportarXivStack layer / Update threadVisual Action Outcome Reasoning Alignment Improves VLM Physical ReasoningarXivStack layer / ContrastDefend agent memory against forged-reasoning poisoning, not just forged facts (FARMA/SENTINEL)arXivStack layer / Follow-up threadCo-LMLM: continuous-query limited-memory language models externalize facts to a knowledge basearXivStack layer / Update threadAnalysis-by-Proxy Finds Latent Localization Signals in VLM Condition EncodersarXiv