NeurIPS 2026 reviews are up: reviews, scores, rebuttals and meta-reviews for 7,877 papers. See the highest rated
acceptodds: the papers of ICLR 2027
309 results for “keyword:"harness"” in ICLR 2027. 64 withdrawn or desk rejected listed apart. clearsort:relevanceratingnewestmost read
[PDF]CoEvo-RSI: A Recursive Self-Improvement Framework for Long-Horizon Model–Harness Co-Evolution▼11k2.33
[PDF]Safe, Persistent, and Evolving Agent Harness for Understanding Partially Observable Worlds▲20k2.67
[PDF]HarnessWeaver: Continual Self-Improvement through Task-Conditioned Harness Composition and Model Routing▼4.3k2.00
[PDF]Forgetting is What to learn: Verifier-Grounded Minimal-Core Skills for Self-Evolving Agents▲3.0k2.33
[PDF]How Well Do Models and Harnesses Work Together: Measuring Model–Harness Fit Beyond Task Scores▼5.8k2.33
[PDF]PhoneHarness: A Mixed-Action Orchestration Harness and Benchmark for Phone Agents across CLI, GUI, and MCP Tools▼4.3k2.00
[PDF]Whose Outcome Is the Team's Score? Auditing Aggregation and Scored-Object Binding in Multi-Agent Evaluation▼12k1.67
[PDF]Divide and Evolve:Orchestrating Agent Harness Evolution with Diagnostic Multi-Agent System▼1.2k1.67
[PDF]MedShard: Task-Structured Inference Improves Clinical Quality Control across Model Scales▼1611.50
Withdrawn or desk rejected (64)
[PDF]HarnessQuotient: Counterexample-Guided Behavioral Quotienting for Self-Evolving Agent HarnessesWithdrawn
[PDF]VideoEvoHarness: Self Evolving Agent Harness from Unlabeled Videos for Video UnderstandingWithdrawn1.75