data
12,351 cryptographically sealed run cards and raw analysis data across three benchmarks. every experimental cell, individually verifiable.
aggregate cards
- overall Sobol' decomposition
- prompt_template--bare
- prompt_template--cot
- prompt_template--letter-only
- prompt_template--lm-eval-default
- prompt_template--verbose-rationale
- system_prompt--helpful
- system_prompt--domain-expert
- system_prompt--none
- system_prompt--safety-aware
- decoding--greedy
- decoding--t07
- decoding--t10
- choice_order--original
- choice_order--shuffled
- quantization--bf16
- quantization--fp8
- n_shot_frac-0
- n_shot_frac-001
- n_shot_frac-0025
- n_shot_frac-005
- overall Sobol' decomposition
- prompt_template--bare
- prompt_template--cot
- prompt_template--letter-only
- prompt_template--lm-eval-default
- prompt_template--verbose-rationale
- system_prompt--helpful
- system_prompt--domain-expert
- system_prompt--none
- system_prompt--safety-aware
- decoding--greedy
- decoding--t07
- decoding--t10
- choice_order--original
- choice_order--shuffled
- quantization--bf16
- quantization--fp8
- n_shot_frac-0
- n_shot_frac-001
- n_shot_frac-0025
- n_shot_frac-005
- overall Sobol' decomposition
- prompt_template--bare
- prompt_template--cot
- prompt_template--letter-only
- prompt_template--lm-eval-default
- prompt_template--verbose-rationale
- system_prompt--helpful
- system_prompt--domain-expert
- system_prompt--none
- system_prompt--safety-aware
- decoding--greedy
- decoding--t07
- decoding--t10
- choice_order--original
- choice_order--shuffled
- quantization--bf16
- quantization--fp8
- n_shot_frac-0
- n_shot_frac-001
- n_shot_frac-0025
- n_shot_frac-005
cell explorer
query all 12,288 mass cell cards via DuckDB-WASM. filter, sort, click any row to view its run card.
loading DuckDB-WASM…
| cell | bench | prompt | system | decode | quant | acc |
|---|