06:00Harness-Delta Attribution: Dissecting Gains From Harness EvolutionResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00ERS: Hierarchical reasoning system based on explicit state graphsResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00ToPA: Block-wise Toeplitz Adaptation for Efficient Fine-TuningResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00When to forget: risk-controlled context adaptation for long-horizon agentsResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Sep-Latent: Separator Tokens for Optimizing Implicit Reasoning in LLMsResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Effective-Horizon Compression in Multi-Turn JailbreaksResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00DiffusionRVGE: Value Gradient Estimation via Regression for Reinforcement Learning in Diffusion ModelsResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00GOLD: Geometric Optimized Latent Diffusion for RNA Structure-Aware Inverse FoldingResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00CPU-GPU collaborative MoE inference system for efficient executionResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00ASE: A modular framework to integrate text in time series forecastingResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Actionable research limitations: forward-citation supervision for automated ideationResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Lost Performance: Analyzing and Fixing C-to-Rust Translation RegressionsResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Distill Until Divergence: Error-Truncated Hindsight Distillation for Agentic RLResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Agentic In-Context Learning: What Algorithms Can Transformers Learn in Closed-Loop Interaction?Research & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00Improving GRPO for safety alignment in large reasoning modelsResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00LexPilot: Benchmarking Proactive Interaction of Legal LLMs with Laypeople in Chinese Civil CasesResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00CIIR: Class-Indexed Interface Representation LearningResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00QuAD: Quantitative assessment of qualitative data quality via distribution alignmentResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Algorithm Modelization: Turning algorithms into evolvable modelsResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00LongForge: Improving long-context instruction data via evidence-assertion reasoning graphsResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00FuzzInsight: Optimizing the Efficiency-Effectiveness Trade-off in Coverage-Guided DNN FuzzingResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Extra test-time compute doesn't always add evidence in LLM reasoningResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00Conceptual framework for analyzing RAG techniquesResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00PrefMind: Online multi-agent preference learning for multimodal response generationResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00Reasoning models control their chains of thought via self-instructionResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00CoTA++: Solving the Repeat Curse in Diffusion-based Multimodal LLMsResearch & Sciencecs.AI updates on arXiv.org9/10Relevance 9 out of 10
06:00SimOPD: Relational Knowledge Transfer for On-Policy DistillationResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00SRCoT: Evidence-Based Text Generation via a Closed Retrieve--Reflect--Revise Loop for Edge LLMsResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00OliO: ODE-based Linear Transition Operator for Self-Supervised Time Series ForecastingResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10
06:00LLM Monitor Validity Failures: Security Evidence Path IssuesResearch & Sciencecs.AI updates on arXiv.org8/10Relevance 8 out of 10