AI Paper Atlas

Research map / Benchmarks / GPQA

GPQA leaderboard

76 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on GPQA. Best accuracy: 96.5 (Strong-Model Reflective Feedback (F3), NeurIPS 2025).

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.

GPQA · accuracy

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

#accuracy ↑MethodPaperTopic
196.5Strong-Model Reflective Feedback (F3)
Llama-4-Maverick
FEEDBACK FRICTION: LLMs Struggle to Fully Incorporate External Feedback
NeurIPS 2025 · text
Agents, code and math
278.9M2CL
Qwen-72B (4 LLMs)
Context Learning for Multi-Agent Discussion
ICLR 2026 · Table 1
LLM agents
378.18SePO-Generalist
Task agent DeepSeek-V3.2, prompt agent Gemini 3.1 Pro Previe
SePO: Self-Evolving Prompt Agent for System Prompt Optimization
NeurIPS 2026 · Table 1
LLM agents
472.1Star Elastic 30B (3.6A)
Nemotron Nano v3 30B
Star Elastic: Many-in-One Reasoning LLMs with Efficient Budget Control
ICML 2026 · Table 1
Reasoning and chain of thought
570.02EELo-CoT
Qwen2.5-32B-Instruct
Activation Control for Efficiently Eliciting Long Chain-of-thought Ability of Language Models
NeurIPS 2025 · Table 3
Reasoning and chain of thought
668.7NAD-kNN
Qwen3-4B-Think
Do LLMs Signal When They’re Right? Evidence from Neuron Agreement
ICML 2026 · Table 1
Reasoning and chain of thought
767.4RRM-32B
DeepSeek-R1-Distill-Qwen-32B
Reward Reasoning Models
NeurIPS 2025 · Binary preference classificati
Reasoning and chain of thought
867.19RLoT (ours)
DeepSeek-R1-Distill-Qwen-7B
RL of Thoughts: Navigating LLM Reasoning with Inference-time Reinforcement Learning
ICLR 2026 · Table 1
Reasoning and chain of thought
967.17ZS-SafePath
R-32B
SAFEPATH: Preventing Harmful Reasoning in Chain-of-Thought via Early Alignment
NeurIPS 2025 · Zero-shot results table
Reasoning and chain of thought
1066.67PAKH
6-LLM pool
Online Multi-LLM Selection via Contextual Bandits Under Unstructured Context Evolution
AAAI 2026 · Table 1
Contextual bandits and Thompson sampling
53 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

GPQA · score

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

#score ↑MethodPaperTopic
132.81LLaDA Base + dLLM-Cache
LLaDA Base
dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching
ICML 2026 · Table 1
Fast diffusion sampling
232.81PRO-B
Mistral-7B-sft
Proximalized Preference Optimization for Diverse Feedback Types: A Decomposed Perspective on DPO
NeurIPS 2025 · Table: Results of aligning Mis
Alignment and preferences
332.8Dream-Instruct + G-Star
Dream-Instruct 7B
Guided Star-Shaped Masked Diffusion
ICML 2026 · Table 1
Fast diffusion sampling
432.8Llama-3.1-8B-Instruct + Paprika
Llama-3.1-8B-Instruct
Training a Generally Curious Agent
ICML 2025 · Table: Evaluation of Paprika o
Human-AI and goal-conditioned
532.8DPO subnetwork finetuning
Llama-3.1-Tulu-3-8B-SFT
Reinforcement Learning Finetunes Small Subnetworks in Large Language Models
NeurIPS 2025 · DPO and PRIME test set perform
RL from feedback
631.91SlowFast Sampling
LLaDA 8B
Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principles
ICLR 2026 · Table 1
Language, safety and interpretability
730.54MPO Iter.4
Llama-3-8B-SFT
Magnetic Preference Optimization: Achieving Last-iterate Convergence for Language Model Alignment
ICLR 2025 · Table: Evaluation results on O
Alignment and preferences
829.78Cal-DPO
Zephyr-7B-SFT
Cal-DPO: Calibrated Direct Preference Optimization for Language Model Alignment
NeurIPS 2024 · Table: UltraFeedback Binarized
Alignment and preferences
929.49XPO-iter3
Llama-3-8B-Flow-SFT
Exploratory Preference Optimization: Harnessing Implicit Q*-Approximation for Sample-Efficient RLHF
ICLR 2025 · Table 1
RL from feedback
1028.9INPO
LLaMA-3-8B SFT
Iterative Nash Policy Optimization: Aligning LLMs with General Preferences via No-Regret Learning
ICLR 2025 · Table 2
RL from feedback
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.