repoGitHubTrust 82 · PrimaryPublished 1mo agoLive · 1mo ago

patrick-toulme/harnessgym

Iterative agent harness improvement: run a coding agent on a hard task, generate the reusable tooling it was missing, qualify it, and replay fresh sessions with it activated. Works with Codex and Claude Code.

Lineage graph

Paper → model → repo connections mined from source citations (Tier-1 exact match).

Why these links exist

Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.

PossiblePossibly related (embedding) · 58%AgentCore-8B →
PossiblePossibly related (embedding) · 52%I built an agent Harness for Small Models. I got Qwen 3.5 4b managing servers. →
PossiblePossibly related (embedding) · 50%Agentic Hardware Design as Repository-Level Code Evolution →
PossiblePossibly related (embedding) · 50%AutoTrainess: Teaching Language Models to Improve Language Models Autonomously →
PossiblePossibly related (embedding) · 50%Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents →
PossiblePossibly related (embedding) · 56%Reasoning effort, not tool access, buys first-try reliability in agentic code generation: an observational study →
PossiblePossibly related (embedding) · 53%CurateEvo: Data-Curation Evolving for Agentic Post-Training →
PossiblePossibly related (embedding) · 65%How self-improving harnesses are rewriting the agent engineering playbook - TechTalks →

Related to

modelAgentCore-8B

Covers

newsI built an agent Harness for Small Models. I got Qwen 3.5 4b managing servers.

Implements

paperAgentic Hardware Design as Repository-Level Code Evolution paperAutoTrainess: Teaching Language Models to Improve Language Models Autonomously paperLearning from Failure: Inference-Time Self-Improvement for Computer-Use Agents

Implements (incoming)

paperReasoning effort, not tool access, buys first-try reliability in agentic code generation: an observational study paperCurateEvo: Data-Curation Evolving for Agentic Post-Training

Covers (incoming)

newsHow self-improving harnesses are rewriting the agent engineering playbook - TechTalks newsA primer on self-improving agent harnesses - Substack newsTraining a harness for model-agnostic and task-environment-agnostic capability improvements with PyTorch-like framework [P]

Related across the graph

newsTraining a harness for model-agnostic and task-environment-agnostic capability improvements with PyTorch-like framework [P]newsA primer on self-improving agent harnesses - Substack paperCurateEvo: Data-Curation Evolving for Agentic Post-Training newsI built an agent Harness for Small Models. I got Qwen 3.5 4b managing servers.newsHow self-improving harnesses are rewriting the agent engineering playbook - TechTalks paperLearning from Failure: Inference-Time Self-Improvement for Computer-Use Agents paperReasoning effort, not tool access, buys first-try reliability in agentic code generation: an observational study modelAgentCore-8B paperAutoTrainess: Teaching Language Models to Improve Language Models Autonomously paperAgentic Hardware Design as Repository-Level Code Evolution

Topics