AI Paper Atlas

Research map / Benchmarks / Minerva Math

Minerva Math leaderboard

27 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on Minerva Math. Best pass@1: 47.4 (DeepCompress-Zero-7B, ICLR 2026).

Numbers are as each paper reports them, read by AI from its results tables and not checked by hand. Papers can differ in setting (training data, model size, extra data); the topic column shows what each paper is about. Open the paper before you compare or quote a number.

Minerva Math · pass@1

14 papers, each with its best reported number for its own method; higher is better.

#pass@1 ↑MethodPaperTopic
147.4DeepCompress-Zero-7B
Qwen2.5-7B
DeepCompress: A Dual Reward Strategy for Dynamically Exploring and Compressing Reasoning Chains
ICLR 2026 · Table 1
Reasoning and chain of thought
247.1SwS-32B
Qwen2.5-32B
SwS: Self-aware Weakness-driven Problem Synthesis in Reinforcement Learning for LLM Reasoning
NeurIPS 2025 · Table 1
Reasoning and chain of thought
345GCPO
DeepSeek-R1-Distill-Qwen-7B
Group Causal Policy Optimization for Post-Training Large Language Models
AAAI 2026 · Table 1
RL from feedback
443.02Dr. MAMR
Qwen2.5-14B-Instruct
Unlocking the Power of Multi-Agent LLM for Reasoning: From Lazy Agents to Deliberation
ICLR 2026 · Table 1
LLM agents
541.3BRiTE-iter-2 ($\psi$-update)
Qwen2.5-7B
BRiTE: Bootstrapping Reinforced Thinking Process to Enhance Language Model Reasoning
ICML 2025 · Table 2
Reasoning and chain of thought
640.4Base + RL (with tool)
Qwen3-4B-Base
GEM: A Gym for Generalist LLMs
ICLR 2026 · Table 1
LLM agents
738.26RankTuner
Qwen3-8B
Probability-Entropy Calibration: An Elastic Indicator for Adaptive Fine-tuning
ICML 2026 · Table 2
Instruction and fine-tuning
831.6TAPO-1.5B
DeepSeek-R1-Distill-Qwen-1.5B
TAPO: Dynamic Teacher and Perturbed Answer Injection for Policy Optimization
AAAI 2026 · Table 1
RL from feedback
931.6GN-IVO (Ours)
Qwen2.5-Math-7B
Group-Normalized Implicit Value Optimization for Language Models
ICLR 2026 · Table 1
RL from feedback
1030.1PS-PPO (Optimized, B=128)
Qwen2.5-Math-7B
PS-PPO : Prefix-Sampling PPO for Critic-Free RLHF
ICML 2026 · Table 2
RL from feedback
4 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Minerva Math · accuracy

13 papers, each with its best reported number for its own method; higher is better.

#accuracy ↑MethodPaperTopic
171.8Ours
Controller Qwen3-4B-Base (System 2: Qwen3-30B-A3B, Ministral
Iterative Critique-and-Routing Controller for Multi-Agent Systems with Heterogeneous LLMs
NeurIPS 2026 · Table 1
LLM agents
250Qwen2.5-72B-Instruct + DAPO
Qwen2.5-72B-Instruct
DAPO : Improving Multi-Step Reasoning Abilities of Large Language Models with Direct Advantage-Based Policy Optimization
NeurIPS 2025 · Mathematical benchmark results
RL from feedback
345.5NestedFP FP8
DeepSeek-R1-Distill-Llama 70B
NestedFP: High-Performance, Memory-Efficient Dual-Precision Floating Point Support for LLMs
NeurIPS 2025 · Table: FP8 Model Quality
Reasoning and chain of thought
437RCD
LLaDA
Residual Context Diffusion Language Models
ICML 2026 · Table 2
Language, safety and interpretability
534.93ReMA
Qwen2.5-7B-Instruct
ReMA: Learning to Meta-Think for LLMs with Multi-agent Reinforcement Learning
NeurIPS 2025 · Table 1
LLM agents
634.56A★-PO
Qwen2.5-7B
Accelerating RL for LLM Reasoning with Optimal Advantage Regression
NeurIPS 2025 · Table 1
RL from feedback
734.2ISLA
Qwen2.5-7B
Beyond Step Pruning: Information Theory Based Step-level Optimization for Self-Refining Large Language Models
AAAI 2026 · Table 1
Alignment and preferences
830.5VeriFree
Qwen3-8B
Reinforcing General Reasoning Without Verifiers
ICLR 2026 · Table 3
Reasoning and chain of thought
925.4TTSV
Qwen3-4B (thinking)
Model Whisper: Steering Vectors Unlock Large Language Models’ Potential in Test-Time
AAAI 2026 · Table 3
Reasoning and chain of thought
1025.3LHM-AIM
R1-Distill-Qwen-1.5B + Qwen2.5-Math-1.5B
Low-Rank Hierarchical Merging for Efficient Long-to-Short Reasoning
NeurIPS 2026 · Table 1
Reasoning and chain of thought
3 more papers on this leaderboard, the baselines reported on it and a CSV download are free with a scime account.
Sign in free to see all

Not affiliated with any conference, journal or the benchmark's authors. Corrections: contact@latenticlabs.com.