AI Paper Atlas

Research map / Benchmarks / MMLU zero-shot

MMLU zero-shot leaderboard

10 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on MMLU zero-shot. Best accuracy: 82.4 (DRAFT-RL, AAAI 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.

MMLU zero-shot · accuracy

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

#accuracy ↑MethodPaperTopic
182.4DRAFT-RL
Claude-3.5-Sonnet
DRAFT-RL: Multi-Agent Chain-of-Draft Reasoning for Reinforcement Learning-Enhanced LLMs
AAAI 2026 · Table 3
LLM agents
279.4QK-score
LLaMA-3.1-70B-Instruct
Listening to the Wise Few: Query–Key Alignment Unlocks Latent Correct Answers in Large Language Models
NeurIPS 2026 · Table 2
Language, safety and interpretability
368.8AlphaAlign
Qwen2.5-7B-Instruct
AlphaAlign: Incentivizing Safety Alignment with Extremely Simplified Reinforcement Learning
ICLR 2026 · Table 3
Reasoning and chain of thought
462.05Spherical Steering
LLaMA-3.1-8B-Instruct
Spherical Steering: Geometry-Aware Activation Rotation for Language Models
ICML 2026 · Table 2
Language, safety and interpretability
555.7SIFT
Llama-33B
Sparse is Enough in Fine-tuning Pre-trained Large Language Models
ICML 2024 · Table 2
Recognition and long-tail
647.1Atlas (de-biased)
Atlas-11B
Atlas: Few-shot Learning with Retrieval Augmented Language Models
JMLR 2023 · Table 6
Retrieval and knowledge
746.2Gradient Cuff
Vicuna-7B-V1.5
Gradient Cuff: Detecting Jailbreak Attacks on Large Language Models by Exploring Refusal Loss Landscapes
NeurIPS 2024 · Table, Robustness-Utility eval
Alignment and preferences
844.1UniAudio 2.0
LLaMA 3.2 3B
A Unified Audio Language Model with Text-Aligned Factorized Audio Tokenization
NeurIPS 2026 · Table 7
Language, safety and interpretability
939.2UL2 20B
UL2
UL2: Unifying Language Learning Paradigms
ICLR 2023 · Table: MMLU 0-shot performance
Language, safety and interpretability
1029.6FIT
Llama-3.1-8B-Instruct
Focus On This, Not That! Steering LLMs with Adaptive Feature Specification
ICML 2025 · Table: Zero-Shot MMLU After FI
Instruction and fine-tuning

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