Hermes agent adapter LoRA for LFM2.5-24B ships with no accuracy loss, +20% efficiency
yb2698 · x · 2026-11-02
yb2698 announced (and Teknium amplified with 👀👀👀) the completion of a Hermes agent adapter LoRA for LiquidAI's LFM2.5-24B, trained for Nous Research's Hermes agent ecosystem. The adapter shows no accuracy drop on held-out test tasks while delivering a 20% efficiency gain, and has been released.
More from coding & agent
- 'duo' skill: two models race blind on the same task, main agent merges results — TheMoonMidas · 2026-11-03
- Aleyda updates Google EEAT content quality checklist with a Claude Skill assessor — gaganghotra_ · 2026-11-03
- Grok Bots testing private sharing: invite users by email, team-wide shared bot browser — XFreeze · 2026-11-03
- Reddit user had Grokbot auto-fill 38 job applications, landing 2 interviews — leebase65 · 2026-11-03
- How we deploy WhatsApp customer agents from one prompt: scoped actions and abuse limits — TheRedfather · 2026-11-03
- Firecrawl launches headless Clay for agents, 88% cheaper, with $75M Series B — gaganghotra_ · 2026-11-03