Good Papers: read the good ones

IAMFlow is a training-free identity-aware memory framework that tracks persistent entities across prompts to generate consistent long narrative videos, achieving best benchmark scores and faster inference.

Jinzhuo Liu, Jiangning Zhang, Wencan Jiang, Yabiao Wang and 4 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 4/5

medium 7/10

strict 1/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026U Hong Kong Southern University of Science a Huazhong University of Science a U Michigan - Ann Arbor Shenzhen Institutes of Advanced Navigation & SLAM

SRL-MPC: Shape-Aware Reinforcement Learned Model Predictive Control

SRL-MPC integrates reinforcement-learned parameter updates with shape-aware model predictive control via geometric separation features to navigate dense heterogeneous robot crowds safely and adaptively.

Ruihua Han, Rui Gao, Zhe Liu, Xinyi WANG and 5 more

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

– 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 5/5

medium 4/10

strict 0/5

83%Must read

?Must readVote to see the score

NeurIPS 2026Tsinghua Huazhong University of Science a Ant International Ant Group Time series forecasting

Existence Precedes Value: Joint Modeling of Observational Existence and Evolving States in Time Series Forecasting

Timeflies jointly infers whether future observations exist and predicts their values, outperforming methods that assume future observation times are known.

Yifan Hu, Hongzhou Chen, Peiyuan Liu, Yiding Liu and 2 more

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

– 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

83%Must read

?Must readVote to see the score

NeurIPS 2026Huazhong University of Science a Alibaba Group Dalian University of Technology Nanyang Technological Chongqing Alignment & preference optimization

Explaining and Breaking the Safety-Helpfulness Ceiling via Preference Dimensional Expansion

MORA rewrites prompts to expand multi-dimensional reward diversity and breaks the safety-helpfulness trade-off, improving sequential single-preference alignment by up to 12.4% and simultaneous alignment by 4.6%.

ShiYing Huang, Liang Lin, Yuer Li, Kaiwen Luo and 5 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 2 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.

80%Must read

?Must readVote to see the score

NeurIPS 2026Fudan Shanghai Jiaotong U Science and Technology of Chin Huazhong University of Science a Shanghai Artificial Intelligence Agent memory

Experience Makes Skillful: Enabling Generalizable Medical Agent Reasoning via Self-Evolving Skill Memory

SkeMex improves medical agents via self-evolving skill memory that distills reusable procedural knowledge, governs retention by utility, and outperforms memory-based agents across clinical tasks.

Haoran Sun, Wenjie Li, Yujie Zhang, Zekai Lin and 7 more

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

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 2026The University of Hong Kong The Chinese University of Hong K Wuhan Huazhong University of Science a Diffusion models

EO-WM: A Physically Informed World Model for Probabilistic Earth Observation Forecasting

EO-WM is a diffusion transformer that forecasts satellite imagery via physically structured weather conditioning and improves vegetation-decline prediction accuracy by up to 7.8% on new diagnostic benchmarks.

Junwei Luo, Shuai Yuan, Zhenya YANG, Yansheng Li and 2 more

Sydney Poster Session 5, Thu, Dec 10, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 3 on Hugging Face · Code ★ 33

– 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

86%Must read

?Must readVote to see the score

NeurIPS 2026Peking Beijing University of Aeronautic Huazhong University of Science a Robot manipulation

SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework

SUGAR converts human videos into humanoid loco-manipulation skills via automated priors, physics refinement, and policy distillation, scaling with video data and enabling zero-shot real-world transfer.

Tianshu Wu, Xiangqi Kong, Yue Chen, Qize Yu and 4 more

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

88%Must read

?Must readVote to see the score

NeurIPS 2026Huazhong University of Science a Institute of automation, Chinese Jilin Institute of Automation, Chinese Carnegie Mellon Video understanding

Rethinking the Readout: Unlocking Video Backbones for AI-Generated Video Detection

Standard video backbone readouts suppress patch-level temporal dynamics needed to detect AI-generated videos; a lightweight velocity-gated patch profiling readout reaches 95.28 AUC on frozen backbones.

Manni Cui, Ziheng Qin, ZiAn Wang, Ruiqi Liu and 7 more

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

– ReadersNo votes yet

15/20 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: 15 of 20 reviewers recommend it

lenient 5/5

medium 9/10

strict 1/5

86%Must read

?Must readVote to see the score

NeurIPS 2026Huazhong University ofscience an Li Auto Inc.Huazhong University of Science a Chinese Academy of Sciences Vision-language models

ITO: Multi-View Alignment and Training-Time Fusion for Image-Text Pretraining

ITO improves image-text pretraining via multi-view cross-modal alignment and discarded training-time fusion, beating CLIP at 100M-1B scale on classification and retrieval.

Hanpeng Liu, Yaqian Li, Zidan Wang, Shuoxi Zhang and 5 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 4/5

medium 8/10

strict 2/5

83%Must read

?Must readVote to see the score

NeurIPS 2026Huazhong University of Science a South China Normal Meituan Alibaba group Autoregressive generation

StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training

StableVQ decouples encoder-decoder and codebook training via Dynamic STE, Region VQ Loss, and independent schedules to stabilize VQ tokenizers and boost utilization and reconstruction.

Bao Tang, Jiahao Guo, Haoxiang Cao, Wenyu Liu and 3 more

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

– 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 2026Huazhong University of Science a Kuaishou- 快手科技 Wuhan Tsinghua Deep RL

TMPO: Trajectory Matching Policy Optimization for Diverse and Efficient Diffusion Alignment

TMPO replaces scalar reward maximization with trajectory-level reward distribution matching via Softmax Trajectory Balance, improving diffusion alignment diversity by 9.1% while avoiding reward hacking and mode collapse.

Jiaming Li, Chenyu Zhu, Zhiyuan Ma, Nanxi Yi and 8 more

Sydney Poster Session 6, Thu, Dec 10, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 3 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.

AI panel: 16 of 20 reviewers recommend it

lenient 5/5

medium 9/10

strict 2/5

74%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Huazhong University of Science a Huazhong University of Science a Xiaomi Corporation Li Auto Inc.Alibaba Group 3D vision & reconstruction

PointForward: Feedforward Driving Reconstruction through Point-Aligned Representations

PointForward reconstructs driving scenes via world-space 3D queries and scene graphs, achieving state-of-the-art feedforward results with explicit cross-view and instance consistency.

Cheng Chi_, Xianqi Wang, Hongcheng Luo, Mingfei Tu and 8 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · 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 2026Huazhong University of Science a Alibaba Group The Hong Kong University of Scie Ant Group Ant Research Video generation

Next Forcing: Causal World Modeling with Multi-Chunk Prediction

Next Forcing uses multi-chunk prediction to accelerate convergence 2.3x, boost high-frame-rate accuracy 93.1%, and double inference speed for world models.

Gangwei Xu, Qihang Zhang, Jiaming Zhou, Xing Zhu and 3 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 8 on Hugging Face · Code ★ 136

– 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 7/10

strict 2/5

83%Must read

?Must readVote to see the score

NeurIPS 2026Mohamed bin Zayed University of Xi'an Jiaotong-Liverpool U Liverpool Huazhong University of Science a The Chinese University of Hong K Medical LLMs

DoAtlas-1: A Causal Compilation Paradigm for Clinical AI

DoAtlas-1 introduces causal compilation to convert medical evidence into executable causal estimands, achieving 98.5% canonicalization accuracy and 80.5% query executability across 1,445 effect kernels.

Yulong Li, Jianxu Chen, Xiwei Liu, Chuanyue Suo and 7 more

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

– 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

86%Must read

?Must readVote to see the score

NeurIPS 2026Huazhong University of Science a Applied Intuition, Inc.Applied Intuition UC Berkeley Hong Kong University of Science Video generation

LaMo: Self-Supervised Latent Motion Priors for Physical Realism in Video Generation

LaMo extracts self-supervised latent motion priors from unlabeled videos via motion drift loss and prior guidance, improving physical consistency in video diffusion without external supervision.

Bo Jiang, Depu Meng, yihan hu, Yichen Xie and 2 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 5/5

medium 9/10

strict 0/5

83%Must read

?Must readVote to see the score

NeurIPS 2026Hong Kong University of Science UC Berkeley Applied Intuition, Inc.U California, Berkeley Huazhong University of Science a Video generation

Teaching Video Generators to Remember: Eliciting Dynamic Memory for Out-of-Sight State Evolution

ReMind trains video diffusion transformers to use cache memory for evolving hidden states across interruptions via memory-oriented curricula and PM-RoPE, achieving best STEVO-Bench scores without catastrophic forgetting.

Tianshuo Xu, Yichen Xie, Depu Meng, Chensheng Peng and 4 more

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

– 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 3/5

medium 9/10

strict 1/5

76%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Ant International Tsinghua Alibaba Group Huazhong University of Science a Department of Computer Science a Time series foundation models

Falcon-X: A Time Series Foundation Model for Heterogeneous Multivariate Modeling

Falcon-X maps heterogeneous time series variates into a unified latent prototype space using diff-attention and latent entity attention to enable cross-variate modeling and zero-shot structural transfer with strong forecasting results.

Yiding Liu, Yifan Hu, Hongjie Xia, Peiyuan Liu and 4 more

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

– 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 3/5

medium 7/10

strict 0/5

89%Must read

?Must readVote to see the score

NeurIPS 2026Huazhong University of Science a Kingsoft Office Chengdu University of Informatio Vision-language models

MDPBench: A Benchmark for Multilingual Document Parsing in Real-World Scenarios

MDPBench introduces a 3,400-image multilingual document parsing benchmark across 17 languages revealing open-source models suffer severe performance drops on photographed and non-Latin script documents.

Zhang Li, Lin Zhibo, Qiang Liu, Ziyang Zhang and 6 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 9 on Hugging Face · Code ★ 895

– 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 8/10

strict 3/5

92%Must read

?Must readVote to see the score

NeurIPS 2026Peking Huazhong University of Science a Institute of Automation, Chinese U Electronic Science and Technol Mohamed Bin Zayed University of LLM evaluation & benchmarks

K12-KGraph: A Curriculum-Aligned Knowledge Graph for Benchmarking and Training Educational LLMs

K12-KGraph introduces a curriculum-aligned K-12 knowledge graph, benchmark, and training data showing current LLMs achieve under 57 percent accuracy on curriculum cognition and that graph-guided supervision outperforms generic instruction tuning.

Hao Liang, Qihan Lin, Mingrui Chen, Hengyi Feng and 6 more

Sydney Poster Session 4, Wed, Dec 9, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 62 on Hugging Face · Code ★ 392

– 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 10/10

strict 4/5

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026OralShanghai Jiao Tong Huazhong University of Science a Tsinghua University, Tsinghua U California, San Diego Carnegie Mellon Unified multimodal models

LatentUM: Unleashing the Potential of Interleaved Cross-Modal Reasoning via a Latent-Space Unified Model

LatentUM unifies modalities in a shared latent space to enable efficient interleaved cross-modal reasoning and generation, achieving state-of-the-art visual planning and self-reflective generation results.

Jiachun Jin, Zetong Zhou, Xiao Yang, Hao Zhang and 3 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 29 on Hugging Face · Code ★ 63

– 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

89%Must read

?Must readVote to see the score

NeurIPS 2026The University of Hong Kong Huazhong University of Science a The Chinese University of Hong K zhejiang university Vision-language-action models

FASTER: Rethinking Real-Time Flow VLAs

FASTER accelerates real-time flow vision-language-action models via horizon-aware sampling that compresses immediate-action denoising into one step, slashing reaction latency on dynamic robot tasks.

Yuxiang Lu, Zhe Liu, Xianzhe Fan, Zhenya YANG and 4 more

Sydney Poster Session 2, Tue, Dec 8, 5:00 PM–8:00 PM, Hall 1-4 · Published 2026 · ▲ 61 on Hugging Face · Code ★ 157

– 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

78%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Wuhan Shanghai Jiao Tong Macquarie Hong Kong University of Science Huazhong University of Science a Efficient inference & serving

Efficient Serving for Dynamic Agent Workflows with Prediction-based KV-Cache Management

PBKV predicts future agent invocations in dynamic LLM workflows to manage KV-cache reuse, achieving up to 1.85x speedup over LRU.

Haoyu Zheng, Fangcheng Fu, Jia Wu, Binhang Yuan and 5 more

Sydney Poster Session 3, Wed, Dec 9, 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 5/5

medium 5/10

strict 1/5

80%Must read

?Must readVote to see the score

NeurIPS 2026City University of Hong Kong Microsoft Huazhong University of Science a Agent memory

M$^\star$: Every Task Deserves Its Own Memory Harness

M* evolves task-specific memory programs via reflective code search to outperform fixed-memory agents across diverse benchmarks. Evolved harnesses develop structurally distinct mechanisms per domain, showing specialization beats general-purpose memory.

Wenbo Pan, Shujie LIU, Xiangyang Zhou, Xianlong Wang and 4 more

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.

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 2026The Hong Kong University of Scie The Chinese University of Hong K Kuaishou Cornell Huazhong University of Science a Deep RL

GARDO: Reinforcing Diffusion Models without Reward Hacking

GARDO selectively regularizes high-uncertainty diffusion samples and adaptively updates reference models to prevent reward hacking while preserving diversity and sample efficiency.

Haoran He, Yuxiao YE, Jie Liu, Jiajun Liang and 7 more

Sydney Poster Session 1, Tue, Dec 8, 10:00 AM–1:00 PM, Hall 1-4 · Published 2026 · ▲ 30 on Hugging Face · Code ★ 63

– 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

71%Highly rated

?Highly ratedVote to see the score

NeurIPS 2026Huazhong University of Science a Physics-informed ML & PDEs

Stable Long-Horizon PDE Forecasting via Latent Structured Spectral Propagators

A structured spectral propagator in a latent space improves long-horizon PDE forecasting stability, reducing relative L2 errors by up to 48.9% versus state-of-the-art methods.

Xiaoxiao Lu, Ye Yuan, Jiahao Shi

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.

AI panel: 7 of 20 reviewers recommend it

lenient 3/5

medium 4/10

strict 0/5