AI Paper Atlas

Research map / Benchmarks / AddSub

AddSub leaderboard

11 papers at ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR report results on AddSub. Best accuracy: 96.71 (CoT + Skill-Based, NeurIPS 2024).

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.

AddSub · accuracy

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

#accuracy ↑MethodPaperTopic
196.71CoT + Skill-Based
GPT-4-0613
Metacognitive Capabilities of LLMs: An Exploration in Mathematical Problem Solving
NeurIPS 2024 · Transfer of Skill Exemplars to
Agents, code and math
294.7PRP (Ours)
text-davinci-003
Get an A in Math: Progressive Rectification Prompting
AAAI 2024 · Table 2
Reasoning and chain of thought
391.9Auto-CoT
Codex code-davinci-002
Automatic Chain of Thought Prompting in Large Language Models
ICLR 2023 · Table: Accuracy using the Code
Reasoning and chain of thought
491.6OccamLlama 8B
Llama 3 8B Instruct
OccamLLM: Fast and Exact Language Model Arithmetic in a Single Step
NeurIPS 2024 · Appendix, reasoning tasks tabl
Agents, code and math
590.55Few-Shot CoT + RCI
gpt-3.5-turbo
Language Models can Solve Computer Tasks
NeurIPS 2023 · Chain-of-Thought comparison ta
Reasoning and chain of thought
688.4Zero-Shot-CoT + TSP + ARI
GPT-3.5-turbo-0125
Rethinking Chain-of-Thought from the Perspective of Self-Training
ICML 2025 · Table 1
Reasoning and chain of thought
787.8Promptbreeder zero-shot
PaLM 2-L
Promptbreeder: Self-Referential Self-Improvement via Prompt Evolution
ICML 2024 · Table 1
Reasoning and chain of thought
881.65STITCH-S
GLM-4-Voice-9B
STITCH: Simultaneous Thinking and Talking with Chunked Reasoning for Spoken Language Models
ICLR 2026 · Table 1a
Reasoning and chain of thought
979.24MoDr
Huginn-3.5B
MoDr: Mixture-of-Depth-Recurrent Transformers for Test-Time Reasoning
ICLR 2026 · Figure 3
Reasoning and chain of thought
1069.6Zero-shot-CoT
text-davinci-002
Large Language Models are Zero-Shot Reasoners
NeurIPS 2022 · Table: Accuracy comparison of
Language, safety and interpretability
1 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.