Search

Any time Past 24 hours Past week Past monthSort:Relevance Newest

Benchmark

MATH-500

Reported in 46 papers on Zotpaper. The latest, each as the paper states it:

Reasoning models can be trained to stop when they know the answer91.8% Accuracy
Language models learn harder problems but can forget easier ones93.24% Pass@128
Parallel training method speeds language model inference by 3.5x4.57 acceptance length

MATH500 (20)All benchmarks →

3 results for “MATH-500”

Research · Language models
LLM pipelines can lose 40.5 accuracy points between stages
The Decomposition Tax: LLM Pipelines Lose Up to 40 Accuracy Points at Their Own Interfaces
Across mathematical reasoning benchmarks, preserving access to the original problem substantially improved multi-stage pipelines without changing the model, prompts, stages or completion budget.
MATH-500: 37 pts accuracy (decomposition tax) (as reported)
29 Sept 2026Rutgers University, Nanjing Tech University, Nanjing University of Posts and Telecommunications
Research · Evaluation & benchmarks
Pass@k misses output diversity changes caused by model post-training
What pass@k Cannot Measure: Evaluating Diversity and Capability Retention after Post-Training
GRPO and rejection-sampling fine-tuning produced sharply different output distributions despite often similar pass@k scores.
GSM8K: 0.63 pass@1 (as reported)
8 Oct 2026Accenture Strategy & Consulting, Vizuara AI Labs
Research · Agents
More Harness Programs Do Not Necessarily Produce Useful Specialization
More Programs or More Rolls? Separating Coverage from Specialization in LLM Harnesses
Controlled comparisons show that apparent gains from diverse LLM harnesses can be matched by repeated runs of identical code.
MATH-500: 98.7% oracle coverage (as reported)
30 Sept 2026The Chinese University of Hong Kong, New Laboratory of Pattern Recognition (NLPR), State Key Laboratory of Multimodal Artificial Intelligence Systems (MAIS)Code