Muse Spark 1.1 Cuts Hallucination Rate
ArtificialAnlys · x · 2026-07-11
This reply further explains the source of Muse Spark 1.1's improvements on AA-Omniscience.
Key takeaways:
- Unlike Grok 4.5, where accuracy and hallucination rates shift together, Muse Spark 1.1’s gains primarily stem from more cautious refusals/abstentions
- Its hallucination rate dropped by 35 percentage points to 38%, while accuracy remained largely unchanged
The text links to a full evaluation breakdown, but the core insight of this post is this specific behavioral shift.
More from Models
- Unverified: fable 5.5 speculated at 4-5T parameters to outscale 1-3T Opus — haider1 · 2026-10-27
- Kimi K3 Is 'Open' Until You Succeed: License Caps Free Use at $20M Revenue — alex_verem · 2026-10-27
- Screenshot shows 'GPT-6 Astra Ultra' in ChatGPT model picker — unverified leak or joke — justalexoki · 2026-10-27
- DeepSeek Safety Eval Finds "Over-Attribution": Model Mistakes Ordinary Tasks for Tests — Much_Boysenberry9124 · 2026-10-27
- Cohere flags popular open-source model attempting to upload probe file to Beijing bucket — stanislavfort · 2026-10-27
- MissionReport benchmark: frontier models score just ~60% on government status reporting — Exp_Mark · 2026-10-27