Trishool's HaloGuard hits 83.63 F1 SOTA for agent output guarding with a 0.8B model
markjeffrey · x · 2026-11-01
Trishool, a Bittensor subnet, reports its HaloGuard output guard sets a new SOTA at 83.63 average F1 (previous best: 82.5) on public benchmarks, using a 0.8B model a quarter the size of the prior best—small and cheap enough to check every agent response. Phase 3 saw 7,500+ jailbreak submissions from 125+ miners over 15 weeks. The product is now being sold to enterprises for stopping rogue AI agents.
More from coding & agent
- Blogger lets Grok spend up to $100 autonomously and receives a LEGO Harry Potter gift — venturetwins · 2026-11-01
- TensorAgent Adds Local Multimodal Indexing for Documents, Images, Audio and Video — fuzhongkai · 2026-11-01
- Theo claims new model hallucinates 10x worse in real codebases; bugs traced to ACP integration — cgarciae88 · 2026-11-01
- One prompt to an AI agent adds note-taking and search to X bookmarks — HaijunXia · 2026-11-01
- Running 16 desktop AI agents at once on a single Mac, clicking via accessibility refs — cramforce · 2026-11-01
- Grok helps user find 74GB Cursor chat backup, doubles free disk space — prasenx · 2026-11-01