Good Papers: read the good ones

Network of Theseus progressively replaces guide network modules with a different target architecture via representational alignment, preserving performance across vastly different deployed architectures.

Vighnesh Subramaniam, Colin Conwell, Boris Katz, Andrei Barbu and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT Diffusion models

Specificity-Aware Diffusion Steering via Variance-Reduced Sequential Monte Carlo

Specificity-aware diffusion steering uses variance-reduced sequential Monte Carlo to suppress undesired regions with minimal positive distribution distortion.

Luran Wang, Linrui Ma, Hannes Stark, Regina Barzilay

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

86%Must read

?Must readVote to see the score

NeurIPS 2026SpotlightMassachusetts Institute of Techn MIT Diffusion models

The Value of Covariance Matching in Gaussian DDPMs and the Lanczos Sampler

Matching full posterior covariance in Gaussian DDPMs reduces path-KL error to O(1/T²), and the matrix-free Lanczos Gaussian sampler achieves this with exponentially decaying approximation error using only Jacobian-vector products.

Sahil Akhtar, Aymane El Gadarri, Vivek Farias, Adam Jozefiak

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026School of Engineering and Applie Harvard MIT Fudan Massachusetts Institute of Techn Agent benchmarks & environments

AI GAMESTORE: Scalable, Open-Ended Evaluation of Machine General Intelligence with Human Games

AI GameStore proposes evaluating general intelligence via scalable synthesis of human games, finding frontier vision-language models score under 10% of human averages on most generated games.

Lance Ying, Ryan Truong, Prafull Sharma, Kaiya Zhao and 8 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 9 on Hugging Face

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

86%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn Shanghai Jiao Tong MIT Molecules & drug discovery

GLACIER: Rethinking Mass Spectrum Prediction as an Object Detection Problem

GLACIER treats tandem mass spectrum prediction as graph object detection, outperforming prior state-of-the-art by up to 19.3% on retrieval accuracy with nearly 8-fold faster inference.

Rui-Xi Wang, Runzhong Wang, Connor Coley

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Anthropic Massachusetts Institute of Techn Transluce UC Berkeley, Transluce AI MIT Mechanistic interpretability

Training Language Models to Explain Their Own Computations

Fine-tuning language models on interpretability ground truth teaches them to describe their internal computations, with self-explanation outperforming larger external explainers.

Belinda Z Li, Zifan Carl Guo, Vincent Huang, Jacob Steinhardt and 1 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 1 on Hugging Face · Code ★ 38

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn Cornell Tech MIT LLM evaluation & benchmarks

The Subjectivity of Monoculture

Monoculture evaluation depends on subjective null-model choices and evaluated model populations, making model agreement a context-dependent inference rather than an absolute property.

Nathanael Jo, Nikhil Garg, Manish Raghavan

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Illinois. Urbana-Champaign U Illinois Urbana-Champaign Carnegie Mellon MIT LLM agents & planning

CTM-AI: A Blueprint for General AI Inspired by a Model of Consciousness

CTM-AI combines a consciousness model with foundation models to integrate diverse processors, achieving state-of-the-art results on multiple benchmarks.

Haofei Yu, Yining Zhao, Lenore Blum, Manuel Blum and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

7/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT EPFL Princeton EPFL - EPF Lausanne Deep learning theory

Provable Quantization with Randomized Hadamard Transform

Dithered randomized Hadamard quantization is unbiased and achieves mean squared error asymptotically matching dense random rotations at O(d log d) cost.

Ying Feng, Piotr Indyk, Michael Kapralov, Dmitrii Krachun and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

83%Must read

?Must readVote to see the score

NeurIPS 2026National University of Singapore Google DeepMind MIT Massachusetts Institute of Techn LLM reasoning & chain of thought

An Enigma of Artificial Reason: Investigating the Production-Evaluation Gap in Large Reasoning Models

LRMs show a large production-evaluation gap, scoring near 48% on reasoning evaluation versus near-perfect production due to answer confirmation bias.

Mingzhong Sun, Teresa Yeo, Armando Solar-Lezama, Tan Zhi-Xuan

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 6 on Hugging Face

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

89%Must read

?Must readVote to see the score

NeurIPS 2026KAUST King Abdullah University of Scie Institute for Computer Science, U Bristol California Academy of Sciences Image recognition

ReefNet: A Large-Scale Dataset and Benchmark for Fine-Grained Coral Reef Recognition

ReefNet provides ~925K genus-level coral annotations and benchmarks showing vision models degrade under zero-shot and cross-source shifts despite adaptation gains.

Abdulwahab Felemban, Yahia Battach, Faizan F Khan, Yuqian Fu and 12 more

Paris Poster Session 2, Wed, Dec 9, 5:00 PM–7:00 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

16/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026MIT MIT CSAIL Facebook NVIDIA Robot manipulation

RigidFormer: Learning Rigid Dynamics using Transformers

RigidFormer is a transformer that learns mesh-free rigid-body dynamics via object-level anchors and differentiable Kabsch projection, outperforming mesh-based baselines with faster inference and scalability to 200+ objects.

Zhiyang Dou, Minghao Guo, Haixu Wu, Doug Roble and 2 more

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 13 on Hugging Face · Code ★ 93

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

86%Must read

?Must readVote to see the score

NeurIPS 2026Harvard Massachusetts Institute of Techn School of Engineering and Applie MIT IIIS, Tsinghua Multi-agent LLM systems

Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions

A decentralized agent economy using auctions and economic selection emerges multi-step reasoning and outperforms monolithic baselines without centralized coordination.

Zhenting Qi, Ao Qu, Huangyuan Su, Chenyu Wang and 12 more

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 10 on Hugging Face · Code ★ 59

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Massachusetts Institute of Techn In-context learning & prompting

LLMs can construct powerful representations and streamline sample-efficient supervised learning

LLMs generate global and local rubrics to standardize multimodal inputs, significantly outperforming clinical baselines on 15 EHRSHOT tasks via sample-efficient supervised learning.

Ilker Demirel, Lawrence Shi, Zeshan Hussain, David Sontag

Atlanta Poster Session 3, Thu, Dec 10, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

83%Must read

?Must readVote to see the score

NeurIPS 2026U California, San Diego Massachusetts Institute of Techn UC San Diego Northwestern Mila; University of Montreal; MI Music

Live Music Diffusion Models: Efficient Fine-Tuning and Post-Training of Interactive Diffusion Music Generators

Live Music Diffusion Models modify diffusion inference with block-wise KV caching to surpass discrete autoregressive efficiency, enabling stable alignment via ARC-Forcing and real-time interactive generation on consumer hardware.

Zachary Novack, Stephen Brade, Haven Kim, Hugo Flores and 5 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 3 on Hugging Face · Code ★ 50

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

91%Must read

?Must readVote to see the score

NeurIPS 2026KU Leuven Department of Electrical Enginee Katholieke Universiteit Leuven KUL Massachusetts Institute of Techn Biosignals & wearables

NeuroAtlas: Benchmarking Foundation Models for Clinical EEG and Brain-Computer Interfaces

NeuroAtlas benchmarks EEG foundation models across 42 datasets and finds they largely match generic time-series models without delivering unified clinical EEG performance.

Konstantinos Kontras, Trui Osselaer, Stylianos G Mouslech, Angeliki I. Karaiskou and 11 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

17/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

89%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn U Toronto MIT Materials & chemistry

Fast Organic Crystal Structure Prediction with Unit Cell Flow Matching

Clari predicts organic crystal structures via unit-cell flow matching with pure pair-bias attention, cutting generation to seconds while surpassing OXtal solve rates and supporting non-sanitizable inputs.

Alston Lo, Luka Mucko, Austin Cheng, Andy Cai and 3 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 2 on Hugging Face

– ReadersNo votes yet

16/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Princeton Massachusetts Institute of Techn Stanford MIT Purdue Multi-agent LLM systems

Improving the Efficiency of Language Agent Teams with Adaptive Task Graphs

LATTE coordinates LLM teams via shared evolving task graphs that reduce tokens, time, and conflicts while matching or exceeding baseline accuracy.

Elizabeth Mieczkowski, Alexander Ku, Tiwalayo Eisape, Dilip Arumugam and 4 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · Code ★ 6

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

91%Must read

?Must readVote to see the score

NeurIPS 2026Princeton Anthropic New York New York University; Anthropic MIT LLM evaluation & benchmarks

DiscoverPhysics: Benchmarking LLMs for out-of-the-box scientific thinking

DiscoverPhysics benchmarks LLM agents on simulated worlds with non-standard physics, finding frontier models pass only half and fail at uncovering latent structure.

Lindsay Smith, Matt Sampson, Siddharth Mishra-Sharma, Peter Melchior and 3 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 1 on Hugging Face

– ReadersNo votes yet

17/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Alberta Massachusetts Institute of Techn MIT U Alberta & Amii Representation & contrastive learning

Measure-to-measure Regression with Transformers

This work formalizes nonlinear measure-to-measure regression and introduces two scalable transformer-based approaches for learning operators between probability distributions. The methods generalize to unseen measures in synthetic experiments, particle systems, and a large-scale colorectal cancer or

Matthew Vandergrift, Martha White, Yury Polyanskiy, Philippe Rigollet and 1 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

7/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026U Edinburgh MIT LLM agents & planning

Autonomous Scientific Discovery via Iterative Meta-Reflection

DiscoPER autonomously discovers scientific patterns via iterative meta-reflection and statistical testing, recovering 8 of 9 ecological patterns and outperforming baselines.

Bingchen Zhao, Sara Beery, Oisin Mac Aodha

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

80%Must read

?Must readVote to see the score

NeurIPS 2026MIT Massachusetts Institute of Techn Optimization

Edge of Stability Selectively Shapes Learning Across the Data Distribution

Edge of stability selectively redistributes learning across data groups via Hessian-aligned gradients and non-vanishing magnitudes, favoring output outliers over saturated ones.

Shauna Kwag, Anakha Ganesh, Tomaso Poggio, Pierfrancesco Beneventano

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Massachusetts Institute of Techn Bayesian & probabilistic methods

To discretize continually: Mean shift interacting particle systems for Bayesian inference

Interacting particle systems extend mean shift to continuous distributions, minimizing maximum mean discrepancy via normalizing-constant-invariant dynamics for fast, multi-modal, high-dimensional quadrature.

Ayoub Belhadji, Daniel Sharp, Youssef Marzouk

Atlanta Poster Session 1, Wed, Dec 9, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

86%Must read

?Must readVote to see the score

NeurIPS 2026U Illinois Urbana-Champaign MIT LLM evaluation & benchmarks

Impacts of Aggregation on Model Diversity and Consumer Utility

Winrate incentivizes model homogenization that reduces consumer welfare, while weighted winrate improves producer specialization incentives and raises utility.

Kate Donahue, Manish Raghavan

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Massachusetts Institute of Techn Domain adaptation

TILT: Target-induced loss tilting under covariate shift

TILT decomposes predictors into main and auxiliary parts, penalizing the latter on unlabeled target data to implicitly weight sources via self-localized, bounded estimands, yielding finite-sample excess risk guarantees and improved domain adaptation performance.

Kakei Yamamoto, Martin Wainwright

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 11 of 20 reviewers recommend it

lenient 3/5

medium 7/10

strict 1/5

86%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT Segmentation

ConnectomeBench2: A Unified Benchmark for Automated Connectomic Proofreading

ConnectomeBench2 unifies multi-species connectomic proofreading data, and a vision transformer trained on it achieves human-level split and merge error correction across species.

Jeff Brown, Tim Farkas, Gleb Razgar, Edward Boyden

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 14 of 20 reviewers recommend it

lenient 5/5

medium 7/10

strict 2/5

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Texas at Austin MIT - Massachusetts Institute of Harvard U California, Berkeley U Texas, Austin LLM reasoning & chain of thought

Interleaved Head Attention

Interleaved Head Attention mixes attention heads via pseudo-heads to enable cross-head reasoning, cutting parameters on synthetic tasks and improving retrieval and math benchmarks over standard multi-head attention.

Sai Surya Duvvuri, Chanakya Ekbote, Rachit Bansal, Rishabh Tiwari and 5 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 2 on Hugging Face

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 11 of 20 reviewers recommend it

lenient 5/5

medium 6/10

strict 0/5

86%Must read

?Must readVote to see the score

NeurIPS 2026SpotlightGenesis Molecular AI MIT Massachusetts Institute of Techn Genesis Therapeutics Genesis Molecular AI Flow matching

Few-Step Cofolding with All-Atom Flow Maps

DeCAF distills all-atom biomolecular cofolding diffusion models into few-step flow maps with SE(3)-aligned endpoint losses, improving accuracy and physical validity at strict inference budgets.

Gianluca Scarpellini, Ron Shprints, Peter Holderrieth, Juno Nam and 6 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 14 of 20 reviewers recommend it

lenient 3/5

medium 10/10

strict 1/5

86%Must read

?Must readVote to see the score

NeurIPS 2026OralHarvard MIT Harvard / MSR LLM pretraining & scaling laws

Bridging Compute- and Data-Optimal Pretraining

A unified compute-data scaling framework introduces token effectiveness to bridge data-rich and fixed-corpus regimes, showing diminishing returns and three operational regimes that render classical compute-optimal allocations suboptimal.

Tian Qin, Kimia Hamidieh, David Alvarez-Melis

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 1 on Hugging Face

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 14 of 20 reviewers recommend it

lenient 5/5

medium 8/10

strict 1/5

83%Must read

?Must readVote to see the score

NeurIPS 2026UMass The University of Tokyo Fudan nvidia U Massachusetts at Amherst Robot manipulation

Action Images: End-to-End Policy Learning via Multiview Video Generation

Action Images formulates robot policy learning as multiview video generation using interpretable pixel-grounded action images, enabling zero-shot control without separate policy heads and improving video-action joint generation.

Haoyu Zhen, Zixian Gao, Qiao Sun, yilin zhao and 6 more

Atlanta Poster Session 2, Wed, Dec 9, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 8 on Hugging Face

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 13 of 20 reviewers recommend it

lenient 5/5

medium 7/10

strict 1/5

83%Must read

?Must readVote to see the score

NeurIPS 2026UIUC Google DeepMind Google Google Research Google Cloud AI Research Agent memory

SkillOS: Learning Skill Curation for Self-Evolving Agents

SkillOS uses RL to train a skill curator that updates an external SkillRepo from experience, improving self-evolving agents across reasoning and multi-turn tasks.

Siru Ouyang, Jun Yan, Yanfei Chen, Rujun Han and 12 more

Sydney Poster Session 3, Wed, Dec 9, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 45 on Hugging Face

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 13 of 20 reviewers recommend it

lenient 5/5

medium 8/10

strict 0/5

80%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT Columbia LLM reasoning & chain of thought

OS-Pruner: Pruning Chains-of-Thought of Reasoning Models via Optimal Stopping

OS-Pruner formulates chain-of-thought pruning as optimal stopping to learn dynamic termination, cutting generation length 20-60% with minimal accuracy loss.

Mohammed Ehab, Aymane El Gadarri, Vivek Farias, Adam Jozefiak and 1 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 12 of 20 reviewers recommend it

lenient 5/5

medium 7/10

strict 0/5

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT Deep learning theory

Learning Orthonormal Bases for Function Spaces

Neural networks parameterize orthonormal function-space bases via ODEs on orthogonal Lie manifolds driven by skew-adjoint generators, with rank-2 generators universally approximating any target basis.

Hamidreza Kamkari, Mohammad S Nabizadeh, Justin Solomon

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 10 of 20 reviewers recommend it

lenient 2/5

medium 5/10

strict 3/5

89%Must read

?Must readVote to see the score

NeurIPS 2026UIUC MIT Massachusetts Institute of Techn LLM pretraining & scaling laws

Training Language Models via Neural Cellular Automata

Neural cellular automata generate synthetic pre-training data that improves language model convergence and downstream reasoning faster than natural text.

Dan Lee, Seungwook Han, Akarsh Kumar, Pulkit Agrawal

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 8 on Hugging Face · Code ★ 88

– ReadersNo votes yet

16/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 16 of 20 reviewers recommend it

lenient 5/5

medium 9/10

strict 2/5

80%Must read

?Must readVote to see the score

NeurIPS 2026SpotlightMIT Google The Institute for Bird Populatio Automated scientific discovery

Targeted Review for AI-Assisted Biodiversity Surveys: Active Continuous-Score Occupancy Modeling

ACORN strategically selects ML predictions for expert review to optimize occupancy model inference, recovering ecological conclusions near fully human-labeled levels with far fewer reviews.

Timm Haucke, Lauren Harrell, Justin Kay, Mary K Clapp and 1 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 12 of 20 reviewers recommend it

lenient 5/5

medium 7/10

strict 0/5

80%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT Fairness & bias

Rank-Constrained Adaptation for Reliable Real-World Performance

MARLA improves worst-group accuracy without subgroup labels by applying a rank-limited logit correction within a low-dimensional misclassification subspace identified from held-out data.

Abinitha Gourabathina, Hyewon Jeong, Teya Bergamaschi, Marzyeh Ghassemi and 1 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 12 of 20 reviewers recommend it

lenient 5/5

medium 6/10

strict 1/5

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Purdue California Institute of Technolo California Institute of Technolo U Illinois Urbana-Champaign Caltech, Asari AI RL for LLMs

Instructing LLMs to Negotiate using Reinforcement Learning with Verifiable Rewards

RLVR trains a 30B LLM buyer via verifiable economic rewards to negotiate, revealing four-phase strategic evolution and outperforming much larger frontier models in surplus extraction.

Shuze D Liu, Claire Chen, Jiabao S Xiao, Lei Lei and 3 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 10 of 20 reviewers recommend it

lenient 5/5

medium 3/10

strict 2/5

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026SpotlightMIT Massachusetts Institute of Techn Speech synthesis

Reconstructing the Vocal Tract with Differentiable Acoustic Simulation

A differentiable GPU acoustic simulator reconstructs vocal tract geometry from speech via gradient descent, enabling cross-lingual autoencoding and unpaired MRI reconstruction.

Eric Chen, Jin Woo Lee, Vincent Sitzmann

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 11 of 20 reviewers recommend it

lenient 5/5

medium 4/10

strict 2/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026The Hebrew University of Jerusal Technion and MIT MIT Learning theory

A Theory of Online Learning with Autoregressive Chain-of-Thought Reasoning

Online autoregressive learning mistake bounds grow from constant to logarithmic in generation horizon M with end-to-end feedback, but chain-of-thought access removes M dependence entirely.

Idan Mehalel, Ilan Doron-Arad, Elchanan Mossel

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 9 of 20 reviewers recommend it

lenient 2/5

medium 4/10

strict 3/5

80%Must read

?Must readVote to see the score

NeurIPS 2026U Pennsylvania U Illinois at Urbana-Champaign New York Massachusetts Institute of Techn MIT LLM reasoning & chain of thought

When Reasoning Meets Its Laws

This paper proposes Laws of Reasoning (LoRe), a framework formalizing reasoning compute and accuracy laws, plus LoRe-Bench showing models lack compositionality; enforcing compute-law compositionality via finetuning improves reasoning performance.

Junyu Zhang, Yifan Sun, Tianang Leng, Jingyan Shen and 3 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 62 on Hugging Face · Code ★ 38

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 12 of 20 reviewers recommend it

lenient 4/5

medium 8/10

strict 0/5

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Massachusetts Institute of Techn Multi-agent RL

GAE Falls Short in Imperfect-Information Self-Play Reinforcement Learning

Standard generalized advantage estimation suffers high variance from stochastic future actions in imperfect-information self-play, so introducing Q-boosting and variance-reduced policy optimization with expected SARSA traces improves performance across large-scale games.

Zhiyuan Fan, Gabriele Farina

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

10/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 10 of 20 reviewers recommend it

lenient 2/5

medium 8/10

strict 0/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT Flow matching

One-Shot Generative Flows: Existence and Obstructions

Straight-line generative flows exist for arbitrary Gaussian endpoints but are impossible for targets with well-separated modes.

Panagiotis Tsimpos, Daniel Sharp, Youssef Marzouk

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 9 of 20 reviewers recommend it

lenient 2/5

medium 5/10

strict 2/5

91%Must read

?Must readVote to see the score

NeurIPS 2026OralMIT Massachusetts Institute of Techn U Wyoming Rostock Fondazione Edmund Mach Spatio-temporal forecasting

MOVEBENCH: A Benchmark for Global-Scale Wildlife Movement Forecasting

MoveBench introduces a 2.6M-location wildlife movement forecasting benchmark across 110 species and finds existing methods generalize poorly to unseen individuals and deep learning does not consistently beat simpler baselines.

Justin Kay, Shir Bar, Ellen O Aikens, Martin Becker and 27 more

Atlanta Poster Session 4, Thu, Dec 10, 4:30 PM–7:30 PM, Hall C1 · Published 2026

100% Readers1 of 1 upvoted

16/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 16 of 20 reviewers recommend it

lenient 5/5

medium 8/10

strict 3/5

83%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn MIT/AWS MIT Long-context modeling

Recursive Language Models

Recursive Language Models let LLMs recursively process long prompts as external environments, handling inputs two orders of magnitude beyond context windows with large quality gains over baselines at comparable cost.

Alex Zhang, Tim Kraska, Omar Khattab

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 100 on Hugging Face · Code ★ 5,673

– ReadersNo votes yet

13/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 13 of 20 reviewers recommend it

lenient 4/5

medium 6/10

strict 3/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Optimization

The Geometry of Linear Program Compression: An Exact Characterization and Learning Algorithm

Geometric characterization and fast-rate learning algorithm exactly compress linear programs into lower-dimensional equivalents preserving optimality with 1/n generalization.

Yuhan Ye, Omar Bennouna

Paris Poster Session 5, Fri, Dec 11, 11:30 AM–1:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 9 of 20 reviewers recommend it

lenient 2/5

medium 6/10

strict 1/5

80%Must read

?Must readVote to see the score

NeurIPS 2026California Institute of Technolo MIT ETH Zurich New York Genesis Therapeutics Molecules & drug discovery

On the Design Space of Discrete Diffusion Online Adaptation for Molecular Optimization

Online discrete diffusion adaptation for molecular optimization finds acquisition, reward shaping, and debiasing complementarily boost reward, with replay and validity control stabilizing exploration to outperform offline and search baselines.

Trevor Chen, Ariel Dai, Jason Yang, Riccardo De Santi and 7 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 12 of 20 reviewers recommend it

lenient 5/5

medium 6/10

strict 1/5

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Massachusetts Institute of Techn Information retrieval

OBLIQ-Bench: Exposing Overlooked Bottlenecks in Modern Retrievers with Latent and Implicit Queries

OBLIQ-Bench reveals modern retrievers miss latent-pattern queries that reasoning LLMs verify once found, exposing a retrieval-verification asymmetry.

Diane Tchuindjo, Devavrat Shah, Omar Khattab

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026 · ▲ 2 on Hugging Face

– ReadersNo votes yet

11/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 11 of 20 reviewers recommend it

lenient 5/5

medium 5/10

strict 1/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Stanford Massachusetts Institute of Techn Standord MIT Optimization

A Single-Sample Polylogarithmic Regret Bound for Nonstationary Online Linear Programming

A re-solving algorithm achieves O(log² n) regret in nonstationary online linear programming using just one sample per distribution via dynamic programming and dual methods.

Haoran Xu, Owen Shen, Peter W Glynn, Yinyu Ye and 1 more

Atlanta Poster Session 5, Fri, Dec 11, 10:00 AM–1:00 PM, Hall C1 · Published 2026

– ReadersNo votes yet

9/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 9 of 20 reviewers recommend it

lenient 4/5

medium 3/10

strict 2/5

86%Must read

?Must readVote to see the score

NeurIPS 2026Massachusetts Institute of Techn Adobe Research Adobe Adobe Research, US MIT Diffusion models

End-to-End Training for Unified Tokenization and Latent Denoising

UNITE unifies tokenization and latent diffusion via a shared generative encoder, enabling single-stage joint training from scratch without adversarial losses or pretrained encoders to reach near state-of-the-art FID scores.

Shivam Duggal, Xingjian Bai, Zongze Wu, Richard Zhang and 4 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026

– ReadersNo votes yet

14/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 14 of 20 reviewers recommend it

lenient 4/5

medium 8/10

strict 2/5

92%Must read

?Must readVote to see the score

NeurIPS 2026Northwestern ScaleAI Scale AI Reflection AI U California, Berkeley Agent benchmarks & environments

MCP-Atlas: A Large-Scale Benchmark for Tool-Use Competency with Real MCP Servers

MCP-Atlas benchmarks LLM tool-use on 1,000 real-server tasks, finding frontier models reach 82.2% pass rates but 63.3% of failures are cognitive.

Chaithanya Bandi, Razvan Dumitru, Ben Hertzberg, Divyansh Agarwal and 15 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026 · ▲ 1 on Hugging Face

– ReadersNo votes yet

19/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 19 of 20 reviewers recommend it

lenient 5/5

medium 9/10

strict 5/5

72%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026MIT Optimization

Parameterized Complexity of Stationarity Testing for Piecewise-Affine Functions and Shallow CNN Losses

Testing approximate stationarity for piecewise-affine functions is XP in fixed dimension but W[1]-hard, with matching lower bounds, extending to shallow CNN losses.

Yuhan Ye

Paris Poster Session 1, Wed, Dec 9, 12:30 PM–2:30 PM, Paris Poster Hall · Published 2026

– ReadersNo votes yet

8/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 8 of 20 reviewers recommend it

lenient 2/5

medium 3/10

strict 3/5

80%Must read

?Must readVote to see the score

NeurIPS 2026School of Computer Science, Carn Carnegie Mellon Nanyang Technological Microsoft Research MIT Explainable AI

Interpreting Neural Combinatorial Optimization via Evolving Programmatic Bottlenecks

EPB distills black-box neural combinatorial optimization models into interpretable program portfolios and reveals stage-dependent heuristic-like behavior.

Haocheng Duan, Yuxin Guo, Jieyi Bi, Anqi Xie and 3 more

Atlanta Poster Session 6, Fri, Dec 11, 4:30 PM–7:30 PM, Hall C1 · Published 2026

– ReadersNo votes yet

12/20 AI panelreviewers recommend it

Readers and the AI panel: vote on this paper to see what they said.

Only vote on papers you've read. Sign in with GitHub to vote.

AI panel: 12 of 20 reviewers recommend it

lenient 5/5

medium 7/10

strict 0/5