ICML 2026 edition · every accepted paper, counted and compared
ICML grew 90% to 6,341 accepted papers. LLM agents grew faster still, from 40 to 281.
LLM agents (Multi-agent systems and games) went from 40 papers at ICML 2025 to 281 this year, from 1.2% to 4.4% of the conference. Inside it, Robust and efficient agent execution grew fastest (3 → 50 papers). Among them: Measuring Agents in Production and RedDebate: Safer Responses Through Multi-Agent Red Teaming Debates.
Also gaining: Reasoning and chain of thought (2.3% → 4.8%) and Multimodal LLMs (1.7% → 4.2%). Losing ground: AI and society (1.9% → 0.6%) and Prediction and decision losses (2.9% → 1.6%).
Qwen is the model family ICML 2026 papers build on most: 1,568 of 5,731 papers with result tables (27%), ahead of Llama (16%) and GPT (OpenAI) (11%). The benchmark named most is GSM8K, then CIFAR-10.
Not everything helped. Of the 69,285 changes the papers measured, 29% made no clear difference or made things worse, as the papers report them. One of 24 on page D:
6,232 papers name an assumption their results go against; 20 of them on page E.
Accepted ICML papers per year in scime atlas.
AI-assisted. Topics come from the atlas clustering; what each paper tested and measured was extracted by AI from its full text. Nothing here was checked by hand and some lines are wrong; open the paper before you cite it.
Describe your own idea in scime atlas to see the closest of 108,967 papers and what similar changes did in them.
Writing it up? scime draft is a free LaTeX editor with AI and the ICML template.
Free for signed-in readers
Five more pages: where the 6,341 papers are, the topics that grew and shrank, what 5,945 papers measured and when it did not help, what the papers say they overturn, and the benchmarks, baselines and models everyone uses. Sign in with a free scime account to keep reading.
Already signed in? Reload this page. Accounts are free; we only use your email to sign you in and, if you agree, to tell you about new editions.
Papers. The 6,341 accepted ICML 2026 papers and 3,342 of 2025 in scime atlas, grouped with all 108,967 atlas papers (ICML, NeurIPS, ICLR, CVPR and AAAI, 2016–2026) into 20 areas and their topics by meaning (an open-source embedding model and clustering; names written with AI).
Full texts. For 6,266 of the 2026 papers an AI model read the full text and recorded the result tables (benchmark, method, base model, number) and each change the paper tested, on what, the metric and whether it helped, plus what the paper says it challenges. These were not checked by hand. Each paper links to its page on scime atlas, which links to the paper itself.
Counting. Shares count each paper once. Citation counts are from OpenAlex and are still low for 2026 papers; papers are ordered by citations, then by how many findings were recorded.
Caveat. Published papers report successes more often than failures. Corrections: contact@latenticlabs.com. Not affiliated with or endorsed by ICML or any of the conferences.
← All editions · © 2026 Latentic Labs