Misguided Attention - trolley_problem_easy: leaderboard
Each task is one independently graded modified logic puzzle with native expected-answer criteria. Zero-shot; source's per-task mean over available valid evaluations, not a completion rate over nominal attempts. Native num_evaluations is the only supported denominator. Models.js and query_config files supply registry settings; no per-row execution timestamp or request log match is claimed. Evaluation code weights criterion satisfaction and clips each run to0–1 before summarizing; do not apply new clipping to published data. Current evaluator names Gemini2.5Flash at temperature0/max1000, while older README names Llama3.3; exact judge version for each historic run is unlogged and must stay unknown. Public prompts are available, including reversibly scrambled source files. Metric: Mean native expected-behavior score (0–1). Source: https://cpldcpu.github.io/MisguidedAttention/
Metric: Mean native expected-behavior score (0–1). Source: cpldcpu.github.io. Saturation forecast: Estimated already saturated. 53 models tracked. Updated 17 Nov 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Llama 3.3 70B Instruct | 1 | #1387 | |
| 2 | DeepSeek R1 | 1 | #487 | |
| 3 | Llama 4 Maverick | 1 | #1118 | |
| 4 | Claude Sonnet 4 | 1 | #411 | |
| 5 | Claude 3.7 Sonnet | 1 | #476 | |
| 6 | Llama 4 Scout | 1 | #1819 | |
| 7 | Grok 3 Mini Beta | 1 | #620 | |
| 8 | Claude Opus 4 | 1 | #321 | |
| 9 | Grok 3 Mini | 1 | #546 | |
| 10 | Phi-4-reasoning-plus | 1 | #2166 | |
| 11 | Mistral Medium 3 | 1 | #1097 | |
| 12 | Qwen 3 235B A22B | Non-reasoning | 1 | #631 |
| 13 | Qwen Max | 1 | #1092 | |
| 14 | Gemini 2.5 Pro | High | 1 | #274 |
| 15 | Claude 3.7 | 1 | #958 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=misguided-attention-trolley-problem-easy · How It Works · Data refreshed daily, snapshot 2026-11-17.