CMU blog: harness engineering boosts scientific agents without retraining models
niloofar_mire · x · 2026-10-03
A CMU Looni Lab team (Suresh Raghu, Kevin Han, Aviral Kumar, Niloofar Mireshghallah) published a blog systematically examining agent harness design — the interface layer around a model controlling tools, state and representation.
Key points:
- Harness engineering can substantially improve agents without retraining the model, and parts of the design process are increasingly automated.
- Coding agents reuse familiar harness components, but scientific agents need domain-specific interfaces, making reuse and automation less obvious.
- On SMDD-Bench (drug-design tasks: lead optimization, scaffold hopping, interaction points) the answer is task-dependent: automated search successfully refined workflows and discovered useful scientific heuristics, while the strongest manual improvements came from changing what evidence the model observes and what state the environment preserves.
- Includes links to the Prime Environment and Harbor Dataset.
Related event: CMU Blog: Automated Harness Search vs Manual Tuning for Science Agents(2 posts)→
More from coding & agent
- New Claude Code plugin renders pasted images as thumbnails above the prompt — jarrodwatts · 2026-10-03
- Coinbase for Agents adds bracket, stop-limit and TWAP orders plus automated feedback loop — kleffew94 · 2026-10-03
- PhantomEnvironments: 7B LLM Trained in Synthetic RL Environments Matches Agents 10x Its Size — CShorten30 · 2026-10-03
- OpenAI introduces dot: cross-app memory, context and Codex task coordination — OpenAIDevs · 2026-10-03
- MIT Interactive Diagrams: From Attention to Mixtral and DeepSeek-V3 Architectures — vtabbott_ · 2026-10-03
- Agent Aware Architecture: an open spec to make sites walkable for AI agents — sierracatalina · 2026-10-03