AtomicChat Ablates Refusals in Liquid AI's LFM2.5, Cutting Them From 96% to 1%
This story is from 2026-11-16. It is preserved in the archive; the latest stories are on the live feed.
AtomicChat released refusal-ablated GGUF builds of Liquid AI's LFM2.5-24B-A2B MoE model, reducing English refusals from 96.3% to 1.2% while slightly improving MMLU scores.
Read the full story at AlphaSignal ↗
Timeline · 1 report
- 2026-11-16 15:29 · AlphaSignal
AtomicChat Strips LFM2.5 Refusals From 96% to 1% by Editing Later Blocks