Latentic LabsPublished by Latentic Labs with scime atlasscime
The scime Dispatch

NeurIPS 2026 edition · every accepted paper, counted and compared

NeurIPS 2026 · 6,080 accepted papersNovember 2026Free · atlas.scime.ai/newspaper

← All editions

LLM Agents: 1.8% of NeurIPS 2025, 5.1% this year

NeurIPS grew 5% to 6,080 accepted papers. LLM agents grew faster still, from 105 to 312.

LLM agents (Multi-agent systems and games) went from 105 papers at NeurIPS 2025 to 312 this year, from 1.8% to 5.1% of the conference. Inside it, Robust and efficient agent execution grew fastest (11 → 57 papers). Among them: Meta-Cognitive Memory Policy Optimization for Long-Horizon LLM Agents and Securing AI Agents with Information-Flow Control.

Also gaining: AI and society (1.0% → 1.9%) and Video generation (1.6% → 2.3%). Losing ground: Agents, code and math (3.4% → 2.6%) and Human and embodied modeling (1.7% → 1.0%).

Qwen is the model family NeurIPS 2026 papers build on most: 922 of 3,424 papers with result tables (27%), ahead of Llama (12%) and GPT (OpenAI) (11%). The benchmark named most is ImageNet, then GSM8K.

Not everything helped. Of the 41,873 changes the papers measured, 31% made no clear difference or made things worse, as the papers report them. One of 24 on page D:

A paper says it overturns

3,785 papers name an assumption their results go against; 20 of them on page E.

Accepted NeurIPS 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.

NeurIPS 2026 in numbers

Accepted papers6,080
NeurIPS 20255,812
Read in full by AI3,811
Findings recorded (change → effect)41,873
Reported as helping65%

Gained most, 2025 → 2026

LLM agents1.8 → 5.1%
AI and society1.0 → 1.9%
Video generation1.6 → 2.3%
Policy gradient and actor-critic0.9 → 1.6%
Robot action and navigation0.8 → 1.5%

Lost most

Agents, code and math3.4 → 2.6%
Human and embodied modeling1.7 → 1.0%
Instruction and fine-tuning1.9 → 1.1%
Molecules and proteins2.8 → 2.1%

Largest areas, 2026

Large language models15.7%
Architectures and efficiency9.1%
Probabilistic and scientific modeling8.9%
Vision-language and multimodal8.8%

Try it on your idea

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 NeurIPS template.

Free for signed-in readers

Read the rest of this edition

Five more pages: where the 6,080 papers are, the topics that grew and shrank, what 3,477 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.

  • B Where the papers are
  • C What grew, what shrank
  • D What the papers measured
  • E What the papers overturn
  • F Benchmarks, baselines and models

Sign in free

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.

G

About the data

how this edition was made

Papers. The 6,080 accepted NeurIPS 2026 papers and 5,812 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 3,811 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 NeurIPS or any of the conferences.

Get the next scime Dispatch and a monthly field update

One email when a new edition is out (ICLR, CVPR, ICML, NeurIPS, ACL and special editions), plus a short monthly note on what moved in machine learning research.

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy

← All editions · © 2026 Latentic Labs