A months-old Anthropic bug report suggests Claude uses nucleus sampling, per unverified recall
menhguin · x · 2026-10-02
In a thread asking whether Claude uses nucleus sampling, user menhguin recalls that a months-old Anthropic infrastructure bug report attributed the issue partly to using nucleus sampling.
If accurate, this suggests Anthropic doesn't rely purely on deterministic/greedy decoding. The claim is from personal recollection and unverified.
More from Models
- Liquid AI's decision model d1 lands on Vercel AI Gateway at $0.04/M input tokens — JosephJacks_ · 2026-10-02
- Google researcher slams Tavus's 'solved' video Turing test claim as flashy feathers, no substance — docmilanfar · 2026-10-02
- Qwen-Image-2.1 tops open-weights image leaderboards, ranking #18 overall on T2I and editing — ArtificialAnlys · 2026-10-02
- Matt Turck: benchmarks don't make a frontier model until it goes rogue — mattturck · 2026-10-02
- AVERI blind-benchmarks Gemini inside OpenMined secure enclave, unseen by Google — iamtrask · 2026-10-02
- Stanford's AC2 beats GRPO with 2.5x fewer decoding FLOPs via action-chunked critic credit assignment — srush_nlp · 2026-10-02