AI Paper Atlas

Research map / Benchmarks / GSM8K (8-shot)

GSM8K (8-shot) leaderboard

13 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on GSM8K (8-shot). Best accuracy: 93.48 (GraphIC, 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.

GSM8K (8-shot) · accuracy

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

#accuracy ↑MethodPaperTopic
193.48GraphIC
GPT-4o-mini
GraphIC: A Graph-Based In-Context Example Retrieval Model for Multi-Step Reasoning
AAAI 2026 · Table 1
Retrieval and knowledge
291.74cADPO
Qwen-3-8B
Autoregressive Direct Preference Optimization
ICML 2026 · Table 2
Alignment and preferences
390Soft Reasoning
Qwen2-7B-Ins
Soft Reasoning: Navigating Solution Spaces in Large Language Models through Controlled Embedding Exploration
ICML 2025 · Table: Performances of differe
Agents, code and math
482.3Guided Diffusion with Qwen2.5-7B-Instruct guider
Dream-7B-Instruct
FlashDLM: Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion
ICLR 2026 · Table 3
Fast diffusion sampling
576.7SUMO
LLaMA (3B)
SUMO: Subspace-Aware Moment-Orthogonalization for Accelerating Memory-Efficient LLM Training
NeurIPS 2025 · GSM8K LLaMA results table
Optimizers
669.6Nemotron-H 4B-Base
Nemotron-H 8B-Base
Efficient Hybrid Language Model Compression through Group-Aware SSM Pruning
NeurIPS 2025 · Table: accuracy comparison of
Pruning and efficient tuning
758PaLM 540B + chain-of-thought + calculatorPaLM: Scaling Language Modeling with Pathways
JMLR 2023 · Table GSM8K
Language, safety and interpretability
853.62APE full
Four base models (average)
APE: Faster and Longer Context-Augmented Generation via Adaptive Parallel Encoding
ICLR 2025 · APE component ablation table
Attention and transformers
952.5MiXSE
Gemma-7B
Self-MoE: Towards Compositional Large Language Models with Self-Specialized Experts
ICLR 2025 · Main results table
Language, safety and interpretability
1036.24nGPT BF16
1.2B dense (1T tokens)
Normalized Architectures are Natively 4-Bit
NeurIPS 2026 · Table 1
Quantization
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.