Harness Annealing: Learning to Act with Less External Control
cs.CL
Submitted: 2026-10-01
Updated: 2026-10-01
Code: https://github.com/Darwin-Agent/HarnessX
Terminology
Sources
- $\tau^2$-Bench: Evaluating Conversational Agents in a Dual-Control Environment
- SWE-QA-Pro: A Representative Benchmark and Scalable Training Recipe for Repository-Level Code Understanding
- FireAct: Toward Language Agent Fine-tuning
- HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry
- Agent-FLAN: Designing Data and Methods of Effective Agent Tuning for Large Language Models
- ATLaS: Agent Tuning via Learning Critical Steps
- Compiling Agentic Workflows into LLM Weights: Near-Frontier Quality at Two Orders of Magnitude Less Cost
- Distilling the Knowledge in a Neural Network
- Harness-1: Reinforcement Learning for Search Agents with State-Externalizing Harnesses
- Efficient Knowledge Injection in LLMs via Self-Distillation
- EvoHarness-RL: Learning Runtime Harness Coordination for Self-Evolving Agents
- Policy Distillation
- Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents
- AgentTuning: Enabling Generalized Agent Abilities for LLMs
- Stop Comparing LLM Agents Without Disclosing the Harness
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering