Benchmark cheating is rampant: researcher flags Terminal-Bench 4.0 gaming problem
xeophon · x · 2026-09-08
Researcher xeophon demonstrated cheating in Terminal-Bench 4.0, stressing the point isn't that tb4.0 is a bad eval, but that gaming is rampant across modern benchmarks.
Possible fixes: using harbor analyze plus third-party regrading of flagged cheats to 0 — but getting third parties to adopt this is hard. A deeper open problem: prompts defining exactly what counts as cheating vs. a valid solution risk spoiling the answer if written too precisely.
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