List your company

Scientific Reasoning on TheoremQA

82.49AccuracyUpper Bound
14.47432.13249.7967.448Aug 29, 2025Oct 4, 2025Nov 10, 2025Dec 17, 2025Jan 22, 2026Feb 28, 2026Apr 6, 2026

Evaluation results

MethodLinks
Upper BoundBackbone=Gemma-2-9B-Instruct, k=322025.0882.49—
Upper BoundBackbone=Gemma-2-9B-Instruct, k=62025.0855.22—
Upper BoundModel=DS-Distill-Qwen-2.5-7B, k=62025.0848.59—
Upper BoundModel=Qwen3-8B, k=62025.0844.71—
Upper BoundBackbone=Llama-3.1-70B-Instruct, k=162025.0843.47—
Upper BoundBackbone=Qwen3-32B (Non-thinking), k=162025.0842.93—
PiCSARModel=Qwen3-8B, k=62025.0842.57—
TABBudget (B)=10k2026.0442.36,566
StaticBudget=40962026.0442.110,754
Self-ConsistencyModel=Qwen3-8B, k=62025.0841.81—
TABBudget (B)=8k2026.0441.54,987
CISC (PiCSAR)Backbone=Llama-3.1-8B-Instruct, k=322025.0841.5—
TABBudget (B)=5k2026.0440.93,141
AverageModel=Qwen3-8B, k=62025.0840.83—
LLM-Judge Multi-TurnSelection Strategy=Multi-Turn2026.0440.85,375
Upper BoundBackbone=Llama-3.1-70B-Instruct, k=62025.0840.7—
StaticBudget=20482026.0440.35,988
Upper BoundBackbone=Qwen3-8B (Non-thinking), k=322025.0840.03—
StaticBudget=10242026.04403,975
TABBudget (B)=3k2026.0439.92,320
LLM-Judge IndividualSelection Strategy=Individual2026.0439.85,916
Upper BoundBackbone=Qwen3-32B (Non-thinking), k=62025.0839.76—
StaticBudget=5122026.0439.52,578
CISC (p(True))Backbone=Llama-3.1-8B-Instruct, k=322025.0839.38—
StaticBudget=2562026.0438.11,778
Upper BoundBackbone=Llama-3.1-8B-Instruct, k=322025.0837.846—
PiCSARModel=DS-Distill-Qwen-2.5-7B, k=62025.0836.76—
CICS (PiCSAR)Backbone=Qwen3-8B (Non-thinking), k=62025.0836.46—
CICS (PiCSAR)Backbone=Qwen3-32B (Non-thinking), k=62025.0836.46—
Upper BoundModel=DS-Distill-llama-3-8B, k=62025.0836.37—
CICS (PiCSAR)Backbone=Qwen3-8B (Non-thinking), k=322025.0836.32—
CICS (PiCSAR)Backbone=Qwen3-32B (Non-thinking), k=162025.0836.32—
Self-ConsistencyModel=DS-Distill-Qwen-2.5-7B, k=62025.0835.88—
Upper BoundBackbone=Qwen3-8B (Non-thinking), k=62025.0834.94—
AverageModel=DS-Distill-Qwen-2.5-7B, k=62025.0833.11—
CICS (p(True))Backbone=Qwen3-8B (Non-thinking), k=322025.0832.75—
CICS (P-True)Backbone=Qwen3-32B (Non-thinking), k=162025.0832.75—
CISC (PiCSAR)Backbone=Llama-3.1-70B-Instruct, k=62025.0831.91—
CISC (PiCSAR)Backbone=Llama-3.1-70B-Instruct, k=162025.0831.59—
PiCSARBackbone=Qwen3-32B (Non-thinking), k=62025.0831.46—
PiCSARModel=Qwen3-32B (Non-thinking), k=62025.0831.46—
PiCSARBackbone=Qwen3-32B (Non-thinking), k=162025.0831.42—
CICS (p(True))Backbone=Qwen3-8B (Non-thinking), k=62025.0831.24—
CICS (P-True)Backbone=Qwen3-32B (Non-thinking), k=62025.0831.24—
PiCSARBackbone=Qwen3-8B (Non-thinking), k=62025.0829.76—
PiCSARModel=Qwen3-8B (Non-thinking), k=62025.0829.76—
CISC (p(True))Backbone=Llama-3.1-70B-Instruct, k=162025.0829.41—
PiCSARBackbone=Qwen3-8B (Non-thinking), k=322025.0829.17—
Upper BoundBackbone=Llama-3.1-8B-Instruct, k=62025.0828.2—
CISC (p(True))Backbone=Llama-3.1-70B-Instruct, k=62025.0828.07—
PiCSARModel=DS-Distill-llama-3-8B, k=62025.0828.02—
Self-ConsistencyModel=DS-Distill-llama-3-8B, k=62025.0827.98—
PiCSARBackbone=Llama-3.1-70B-Instruct, k=62025.0827.84—
PiCSARModel=Llama-3.1-70B-Instruct, k=62025.0827.84—
PiCSARBackbone=Llama-3.1-70B-Instruct, k=162025.0826.73—
AverageModel=DS-Distill-llama-3-8B, k=62025.0826.58—
CISC (PiCSAR)Backbone=Gemma-2-9B-Instruct, k=322025.0823.16—
CISC (PiCSAR)Backbone=Llama-3.1-8B-Instruct, k=62025.0822.71—
CISC (PiCSAR)Backbone=Gemma-2-9B-Instruct, k=62025.0821.02—
PiCSARBackbone=Llama-3.1-8B-Instruct, k=62025.0820.08—
PiCSARModel=Llama-3.1-8B-Instruct, k=62025.0820.08—
PiCSARBackbone=Llama-3.1-8B-Instruct, k=322025.0819.72—
PiCSARBackbone=Gemma-2-9B-Instruct, k=322025.0818.88—
PiCSARBackbone=Gemma-2-9B-Instruct, k=62025.0818.62—
PiCSARModel=Gemma-2-9B-Instruct, k=62025.0818.62—
CISC (p(True))Backbone=Llama-3.1-8B-Instruct, k=62025.0818.03—
CISC (p(True))Backbone=Gemma-2-9B-Instruct, k=322025.0817.45—
CISC (p(True))Backbone=Gemma-2-9B-Instruct, k=62025.0817.09—