AI Paper Atlas

Research map / Benchmarks / HellaSwag (zero-shot)

HellaSwag (zero-shot) leaderboard

10 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on HellaSwag (zero-shot). Best accuracy: 86.7 (QK-score, NeurIPS 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.

HellaSwag (zero-shot) · accuracy

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

#accuracy ↑MethodPaperTopic
186.7QK-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
280.8ChinchillaAn empirical analysis of compute-optimal large language model training
NeurIPS 2022 · Table 5
Pruning and efficient tuning
3726.7B+muP
GPT-3 6.7B
Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer
NeurIPS 2021 · Table 7
Pruning and efficient tuning
460.9OmniMoE
6.4B-A1.7B (from scratch)
OmniMoE: An Efficient MoE by Orchestrating Atomic Experts at Scale
ICML 2026 · Table 1
Mixture of experts
551.83MDM (1.1B)Scaling up Masked Diffusion Models on Text
ICLR 2025 · Language understanding results
Fast diffusion sampling
642.37InfoDLM
1B
InfoDLM: an Information-Adaptive Framework for Discrete Diffusion Language Model Pretraining
ICML 2026 · Table 2
Reasoning and chain of thought
738.5MoSE (TTT)
GPT2-Medium
MoSE: Mixture of Slimmable Experts for Efficient and Adaptive Language Models
ICML 2026 · Table 2
Mixture of experts
836SALAAD eL+eS (646M)
LLaMA-based 1B
SALAAD: Sparse And Low-Rank Adaptation via ADMM for Large Language Model Inference
ICML 2026 · Table 2
Pruning and efficient tuning
933.2Leviathan
1.2B decoder
Leviathan: Decoupling Input and Output Representations in Language Models
NeurIPS 2026 · Table 2
Attention and transformers
1027.5LocalRegret
OLMoE-1B-7B, 4B tokens from scratch
Regret Pre-training: Bridging Prior and Posterior Views for Enhanced Knowledge Grounding
ICML 2026 · Table 1
Knowledge distillation

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