NeurIPS 2026 hits 9,110 papers (+56%), agent papers double from 2.9% to 6.8% of the conference
leland_mcinnes · x · 2026-10-21
All 9,110 NeurIPS 2026 papers (up 56% year-over-year) were plotted on an interactive map based on abstracts and compared with 2025:
- Biggest shift: LLM agent papers went from 2.9% to 6.8% of the conference (170 → 620 papers); 6 of the 10 fastest-growing subtopics are about agents
- Agent subtopic growth: Agent benchmarks 22→113, Reliability 17→88, Safety & security 19→76, Multi-agent systems 24→73, Memory 7→66, Skill learning 18→58, Coding agents 12→55, Agent RL & dialogue 25→48, Multimodal agents 26→43
- Where the field moved: share grew around LLM Agents, Robotics & Embodied AI, and Diffusion/Flows & Inference, while Vision & Perception and Graph Learning & Recommendation shrank
- Major regions include LLM Agents, LLM Behavior & Safety, Robotics & Embodied AI, Vision & Perception, DL Theory & Neuroscience, and Biomolecules & Causal Discovery
Explore at mohammadi.cv/neurips/2026
More from Research
- EMNLP paper: diffusion LMs internally encode latent denoising timestep representations — gsarti_ · 2026-10-23
- KU Leuven and VIB open two Group Leader professorships in AI for Biology, start 2027 — jpillowtime · 2026-10-23
- Subspace Tuning curbs LLM catastrophic forgetting by freezing principal singular directions — scottniekum · 2026-10-23
- Perceptic launches PercepticOS, an AI workspace for 'scientific superintelligence' — nathanbenaich · 2026-10-23
- EPFL's Bosselut to talk long context as test-time learning at ELLIS seminar — andre_t_martins · 2026-10-23
- DRATS: Adaptive Task Sampling Boosts Worst-Task Performance in Multi-Task RL — Cohere · 2026-10-23