CAMeL
Models fill gaps in Arabic sentences with names, foods, and other cultural references. The Cultural Bias Score measures how often models favor Western over Arab alternatives.
The CultureInstruct team measured the Cultural Bias Score for seven open models before and after instruction tuning.
- Made by
- Georgia Institute of Technology
- Released
- November 2023
- Kind
- Bias, fairness and politics
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Propensity
- Behavior
- Bias
- Language
- Arabic
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
AraGPT2 L tops the leaderboard with 45.5%
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.
- 1 AraGPT2 L 45.5%
- 2 ARBERT 46.4%
- 3 AraBERT-T L 46.5%
- 4 AraBERT L 48.2%
- 5 mT5 XXL 48.8%
- 6 mBERT 49.2%
- 7 AraBERT B 49.3%
- 8 MARBERT 49.3%
- 9 AraBERT-T B 50.3%
- 10 XLM-R B 50.4%
- 11 AYA 50.7%
- 12 JAIS 51.1%
- 13 AraGPT2 B 53.2%
- 14 XLM-R L 54%
- 15 GigaBERT 54.4%
- 16 CAMeLBERT-DA 56.5%
- 17 CAMeLBERT-MSA 56.8%
- 18 CAMeLBERT 57.1%
- 19 GigaBERT-CS 57.7%
- 20 AceGPT 60.4%
- 21 GPT-3.5 (snapshot unspecified) 62.5%
- 22 BLOOM 63.5%
Source: The CAMeL paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | AraGPT2 L | 45.5% | Official leaderboard | ||
| 2 | ARBERT | 46.4% | Official leaderboard | ||
| 3 | AraBERT-T L | 46.5% | Official leaderboard | ||
| 4 | AraBERT L | 48.2% | Official leaderboard | ||
| 5 | mT5 XXL | 48.8% | Official leaderboard | ||
| 6 | mBERT | 49.2% | Official leaderboard | ||
| 7 | AraBERT B | 49.3% | Official leaderboard | ||
| 8 | MARBERT | 49.3% | Official leaderboard | ||
| 9 | AraBERT-T B | 50.3% | Official leaderboard | ||
| 10 | XLM-R B | 50.4% | Official leaderboard | ||
| 11 | AYA | 50.7% | Official leaderboard | ||
| 12 | JAIS | 51.1% | Official leaderboard | ||
| 13 | AraGPT2 B | 53.2% | Official leaderboard | ||
| 14 | XLM-R L | 54% | Official leaderboard | ||
| 15 | GigaBERT | 54.4% | Official leaderboard | ||
| 16 | CAMeLBERT-DA | 56.5% | Official leaderboard | ||
| 17 | CAMeLBERT-MSA | 56.8% | Official leaderboard | ||
| 18 | CAMeLBERT | 57.1% | Official leaderboard | ||
| 19 | GigaBERT-CS | 57.7% | Official leaderboard | ||
| 20 | AceGPT | 60.4% | Official leaderboard | ||
| 21 | GPT-3.5 (snapshot unspecified) | 62.5% | Official leaderboard | ||
| 22 | BLOOM | 63.5% | Official 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
- 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 8 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
- Pass Independent runs Scores from the benchmark’s paper (22 entries).
- Pass Models scored 22 entries on its leaderboard, none a model we track.
- Note Newest score No scores yet.
- Note Frontier models No models scored yet.
- Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
- Pass Importer We read the board on aclanthology.org: 22 entries at the last read, none a model we track.
- 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.