Tag
Open-source AI tooling
1 published trend · vote, save or discuss any of them
- 01AI Workflows & AutomationEmerging7D: Insufficient history
Hugging Face trains harness adapters for agent tasks
Hugging Face introduced reusable harness adapters that apply reinforcement learning inside a model’s agent harness. Hugging Face reports that a LoRA adapter for Qwen3.8-27B raised Terminal-Bench 2.1 from 39% to 55% pass@1 under Pi while reducing average token use from about 2.9K to 1.5K per task.
Published
Frontier & foundation modelsAI agents & automationOpen-source & local AI51 out of 100- 7D
- +100%
- Signals
- 1
- Indep. src
- 1
- Confidence
- Unconfirmed