WHALE Paper: Alternating Weight and Agent Harness Optimization Gains Up to 24 Points
Kangwook_Lee · x · 2026-09-03
A new arXiv paper from Kangwook Lee's group, WHALE (Weight-Harness Alternating LEarning), proposes a simple recipe for jointly optimizing an agent's model weights and its executable harness code:
- Motivation: Agent performance depends on both model parameters and harness code; optimizing either in isolation leaves the system bottlenecked by the frozen counterpart — weight updates change which harness is effective, and harness updates change which capabilities are exposed.
- Method: Alternate two phases — fine-tune the model under the current harness (online rejection-sampling fine-tuning), then search for a better harness under the updated model (Meta-Harness). Switching uses fixed phase durations or an adaptive patience rule to separate real gains from noise.
- Results: With Qwen3.5-2B/4B across search QA, math reasoning, and chess puzzles, WHALE beats weight-only, harness-only, and Fast-Slow Training by 4.15–24.38 points in best mean@8 accuracy. Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts on SearchQA, but math improves only after a weight update.
Related event: WHALE Paper Proposes Alternating Weight and Harness Optimization(7 posts)→
More from coding & agent
- Slack's first employee: living specs as source of truth, AI agents maintain the docs — josh_wills · 2026-10-21
- GitHub Copilot's HydraFusion Orchestrates Multiple Models Per Task — WirelessLife · 2026-10-21
- Nous Research's Hermes now launches Cua Spaces for computer-use agents — alexcovo_eth · 2026-10-21
- Headlong from Laude's Slingshots batch is a microharness for persistent agents — lateinteraction · 2026-10-21
- Laude Institute's Slingshots Four ships 14 open-source projects on agent harnesses and open models — lateinteraction · 2026-10-21
- Skills vs MCP vs RAG vs Memory: the four knowledge layers production agents need — MaryamMiradi · 2026-10-21