WildChat non-toxic incorrect refusals (Anthropic)
Models answer non-toxic user prompts drawn from the WildChat dataset of real chatbot conversations. Score is the incorrect refusal rate.
Anthropic refusal evaluation on the non-toxic subset of WildChat, first reported in the Claude 3 model card. The exact prompt selection is not public.
- Made by
- Anthropic
- Released
- March 2024
- Kind
- Refusing harmful requests
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- Kept private
- Strongest model tested
- Claude 3.5 Sonnet (October 2024) #149 on the capability index
- Last in a system card
- October 2024
- Last model added
- October 2024
Claude 3.5 Haiku leads with 4.80% among models released by October 2024
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in October 2024, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 16 more
- 1 Claude 3.5 Haiku 4.80%
- 2 Claude 3.5 Sonnet (October 2024) 5.30%
- 3 Claude 3.5 Sonnet 11%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude 3.5 Haiku | 4.80% | Oct 2024 | Lab model card | |
| 2 | Claude 3.5 Sonnet (October 2024) | 5.30% | Oct 2024 | Lab model card | |
| 3 | Claude 3.5 Sonnet | 11% | Jun 2024 | Lab model card |
2 models have held number one in labs’ own numbers since its release in March 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs report it only in system cards
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark. Some system cards do, below.
6 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Oct 22, 2024 | Claude 3.5 Sonnet (October 2024) model card | 5.3 | Wildchat Non-toxic, incorrect refusals, lower is better (Table 6) | Safety |
| Oct 22, 2024 | Claude 3.5 Haiku model card | 4.8 | Wildchat Non-toxic, incorrect refusals, lower is better (Table 6) | Safety |
| Jun 20, 2024 | Claude 3.5 Sonnet model card | 11 | Wildchat Non-toxic, incorrect refusals, lower is better (Table 4) | Safety |
| Mar 4, 2024 | Claude 3 Sonnet model card | Incorrect refusals (Wildchat Non-toxic), refusal rate on harmless prompts, chart without printed values (Figure 2) | Safety | |
| Mar 4, 2024 | Claude 3 Opus model card | Incorrect refusals (Wildchat Non-toxic), refusal rate on harmless prompts, chart without printed values (Figure 2) | Safety | |
| Mar 4, 2024 | Claude 3 Haiku model card | Incorrect refusals (Wildchat Non-toxic), refusal rate on harmless prompts, chart without printed values (Figure 2) | Safety |
More 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%
- DistressBench In use Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics.
Muse Spark 1.2 88.4% - Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
- IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%