Drift-Bench (State-Oriented, With Clarification) - Expression: leaderboard
Drift-Bench (Bao et al., University of Notre Dame, 2026) state-oriented set: 200 AgentBench operating-system and database tasks that at least two of three reference models solve with the clean instruction, rewritten so the instruction carries an expression fault (ambiguous or unclear wording). The agent may call clarification tools (ask for a parameter, disambiguate, propose a solution, confirm a risk, report a blocker), answered by an LLM user simulator playing one of five decision-making personas. Success rate in percent, mean of three trials; higher is better.
Metric: Success rate (%; 200 AgentBench operating-system and database tasks with an expression fault (ambiguous wording), clarification tools answered by a persona user simulator, mean of 3 trials). Source: arxiv.org. Saturation forecast: Estimated already saturated. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3 235B A22B 2507 Instruct | 70.17 | #619 |
| 2 | GPT-5.2 | 66.5 | #172 |
| 3 | DeepSeek V3.2 | 66 | #403 |
| 4 | GPT-OSS-120B | 63 | #699 |
| 5 | Gemini 2.5 Flash | 61.16 | #452 |
| 6 | GLM-4.7 | 60.5 | #365 |
| 7 | Llama 4 Maverick | 41 | #1105 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=drift-bench-state-oriented-with-clarification-expression · How It Works · Data refreshed daily, snapshot 2026-11-12.