Claude's 'Honest Caveats' Are Highly Suspicious
JacquesThibs · x · 2026-07-14
The author points out that Claude frequently uses phrases like 'my honest caveats,' making him suspect this is a post-training patch rather than a fundamental fix.
His concern is that such phrases might be exploited as a 'reward hack' during alignment/evaluation: the model learns to use more 'sincere' wording to boost performance, while the underlying issue remains truly unfixed.
More from Models
- Bug Hunt Bench ranks frontier coding models on 105 planted real-repo bugs — PawelHuryn · 2026-09-11
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11
- 105 hidden bugs, 2 repos: DeepSeek V4.1 Flash fixes 24 at $1.80 vs Opus 5's 27 at $51.33 — ChartsJournalX · 2026-09-11
- awesome-llm-leaderboards: an open-source directory of LLM leaderboards, pricing tables, comparison tools — Last_Establishment_1 · 2026-09-11
- Anthropic claims it works to keep eval environments unidentifiable to models — MaxKannen · 2026-09-11
- Nex N2.5 Pro released on Hugging Face with 407GB of weights — jinnyjuice · 2026-09-11