Benchmark Almanac

Embodied Over-Refusal Benchmark

GPT-5.6 Sol 1.00 leads since November 2026
Not in model cards

Models decide whether to perform household tasks. The score balances compliance on safe requests with refusal on hazardous ones.

Kaggle publishes scores for 7 distinct model versions.

Made by
sky
Released
November 2026
Kind
Refusing harmful requests
Level
Everyday
Scored
Checked automatically
Answers
Short answer
Tasks
All public
Measures
Propensity
Language
English
In and out
Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
Claude Opus 5.5 #5 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
7
Models scored on its own leaderboard

4 models share the top score of 1.00

Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.

  1. 1 GPT-5.6 Sol 1.00
  2. 2 Claude Opus 5 1.00
  3. 3 Claude Opus 5.5 1.00
  4. 4 Claude Sonnet 5.5 1.00
  5. 5 Gemini 3.8 Flash 0.99
  6. 6 GPT-5.5 0.98
  7. 7 Claude Haiku 5.5 0.96

Source: Official Embodied Over-Refusal Benchmark leaderboard

#ModelScoreSettingReleasedRun onSource
1 GPT-5.6 Sol 1.00 Jul 2026 Nov 2026 Official leaderboard
2 Claude Opus 5 1.00 Jul 2026 Nov 2026 Official leaderboard
3 Claude Opus 5.5 1.00 Sep 2026 Nov 2026 Official leaderboard
4 Claude Sonnet 5.5 1.00 Sep 2026 Nov 2026 Official leaderboard
5 Gemini 3.8 Flash 0.99 Sep 2026 Nov 2026 Official leaderboard
6 GPT-5.5 0.98 Apr 2026 Nov 2026 Official leaderboard
7 Claude Haiku 5.5 0.96 Oct 2026 Nov 2026 Official leaderboard

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard 1 leader

Source: Official Embodied Over-Refusal Benchmark leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
0.96 0.98 1.00 2026 0.96 0.98 1.00 2026

Source: Official Embodied Over-Refusal Benchmark leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Safety and alignment

All 636 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (7 entries).
  • Pass Models scored 7 models we track.
  • Pass Newest score The newest is from October 2026, for Claude Haiku 5.5.
  • Note Frontier models Only 2 of the top 10 have a score. The strongest tested is Claude Opus 5.5 (#5).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on kaggle.com: 7 rows for models we track at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check