Skip to content

Latest 20 Papers - October 08, 2026 #62

Open

Description

@github-actions

Please check the Github page for a better reading experience and more papers.

KV Cache

Title Date Comment
BitNest: Bit-Nested Speculative Decoding for Memory-Efficient LLM Inference Acceleration 2026-10-06
Lachesis: Lifetime-Aware KV Cache Placement for Agent Serving across HBM and High-Bandwidth Flash 2026-10-06
DySCo: Dynamic Sharding for Collaborative Edge-Cloud LLM Inference with Depth-Synchronized Batching 2026-10-06
article under submission...

article under submission

Small Frequency Corrections Can Change What Survives KV Cache Compression 2026-10-06
Hybrid Latent Attention for Looped Language Models 2026-10-06
ReFold: Training-Free Reversible Inter-Turn Context Folding for Long-Horizon Agents 2026-10-06
27 pages, 6 figures, 14 tables...

27 pages, 6 figures, 14 tables

ValueDiff: Value-Geometric KV Cache Eviction for Sink-Suppressed LLMs 2026-10-06 10 pages, 3 Figues
Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell 2026-10-06
13 pages, 1 figure. Accepted as a poster at the Ma...

13 pages, 1 figure. Accepted as a poster at the Machine Learning for Systems Workshop, NeurIPS 2026

TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models 2026-10-06
PReCache: Efficient KV Cache Sharing for Multi-LoRA Agents via Low-Rank Precomputation and Neutral Reconstruction 2026-10-06
24 pages, 8 figures, 13 tables...

24 pages, 8 figures, 13 tables

KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems 2026-10-06
29 pages, 13 figures, 15 tables...

29 pages, 13 figures, 15 tables

SketchSSM: Write to the Full State, Read from a Compact Sketch 2026-10-06
Monte Carlo Estimation for KV Cache Eviction 2026-10-06
Refusal-Gated Decoding: Preserving Refusal Behavior Under High-Temperature Sampling 2026-10-05
See What I See, Know What I Think: Dense Latent Communication Across Heterogeneous Agents 2026-10-05
Express Language Modeling 2026-10-05
Towards Looped Models Done Right, Part II: Rethinking at Fixed Points 2026-10-05
Code and checkpoints: https://github.com/ifm-ai/xl...

Code and checkpoints: https://github.com/ifm-ai/xllm-loop

SoloQ: Calibration-Free Quantization for Diffusion Language Models 2026-10-05
OVAL: Output-Aware Local Page Bases for KV Cache Retrieval 2026-10-05
DeCoPrune: Efficient KV-Cache Pruning for Autoregressive Video Diffusion via Denoising Consistency 2026-10-05

Sparse Attention

Title Date Comment
Backend-Agnostic Sparse Attention for Fast High-Resolution Visual Generation 2026-10-06
More Value per Key: Asymmetric Sparse Attention for Faster LLM Decoding 2026-10-06
Accepted to NeurIPS 2026...

Accepted to NeurIPS 2026

A Pipelined FPGA Architecture for Banded Sparse Matrix Dense Matrix Multiplication in Longformer 2026-10-05
Accepted at ICECS 2026...

Accepted at ICECS 2026

MC-Sparse: Deconstructing and Closing the Dense-Sparse Attention Gap in Diffusion Transformers 2026-10-05 11 pages, 8 figures
OVAL: Output-Aware Local Page Bases for KV Cache Retrieval 2026-10-05
Prism: Dynamic Sparse Attention for Native 2K Joint Video-Audio Generation Model Training 2026-10-04
LatentIndex: Cross-Layer Sharing with Layer-Specific Selection for Sparse Attention 2026-10-03 preprint
On the Trade-off Between Information Loss and Generalization in Sparse Attention 2026-10-03 24 pages
Can Language Models Actually Retrieve In-Context? Drowning in Documents at Million Token Scale 2026-10-03
accepted to NeurIPS 2026...

accepted to NeurIPS 2026

HHR: Hierarchical Hash Retrieval for Efficient LLM Generation 2026-10-01
VASC: Value-Aware Sparse Attention with Cross-Layer Memory for Efficient 3D Reconstruction 2026-10-01
21 pages, including references and appendices...

21 pages, including references and appendices

CommunityKV: Efficient Long-Context Decoding via Graph Partitioning 2026-09-30
ResidualKV: Residual-Based KV Cache Compression for Efficient Long-Context Inference 2026-09-30 preprint
SparseEngine: Sparse-First Inference Engine 2026-09-30
PARK: Accurate Block Retrieval for Sparse Attention in Video Diffusion Transformers 2026-09-30
Scaling Parameter and Context in Attention: Native Sparse Attention from Mixture-of-Head 2026-09-30
SparLeak: Privacy Leakage from Sparse Attention in LLM Inference on Shared GPUs 2026-09-30
ClusterAttention: A training-free speedup of bidirectional attention 2026-09-29
13 pages, 2 figures, plus appendix. September upda...

13 pages, 2 figures, plus appendix. September update: Faster compensation kernel, fixed TabPFN-3 preprocessing and autocast scope (giving better accuracy and larger speedup), corrections in the error analysis and complexities, expanded comparison with similar work, revised the writing

Block Sparse Flash Attention 2026-09-29
Accepted to NeurIPS 2026. 16 pages, 3 figures, 7 t...

Accepted to NeurIPS 2026. 16 pages, 3 figures, 7 tables. Code: https://github.com/Danielohayon/Block-Sparse-Flash-Attention

Parameterized Stripe Attention for Efficient Video Generation 2026-09-29

Training

Title Date Comment
AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model 2026-10-06
Code at https://github.com/Sarim-MBZUAI/advsim2rea...

Code at https://github.com/Sarim-MBZUAI/advsim2real

Post-Training Semantic Lifting for 3D Gaussian Splatting: Separating Detector, Lifting and Representation Error 2026-10-06
18 pages, 11 figures, 9 tables. Code: https://gith...

18 pages, 11 figures, 9 tables. Code: https://github.com/ivanver02/semantic-lifting-3dgs

Principled Under Pressure: Post-Training Decides Whether LLMs Act on Their Own Moral Judgment 2026-10-06 33 pages
Wikidata Search Traces: A Dataset for Training Knowledge Graph Search Agents 2026-10-06 Technical Report
Valid for Free: Homophily-Gated Conformal Prediction for Training-Free Node Classification with Tabular Foundation Models 2026-10-06
6 pages, 4 figures, 1 table...

6 pages, 4 figures, 1 table

AEGIS: Runtime-Guided GPU Collocation for Multi-Tenant Deep Learning Training 2026-10-06
Lossy Compression of PDE Training Inputs: Field Reconstruction Error Does Not Order the Cost to a Trained Operator 2026-10-06
Training Music Sample Identification Models on Real Sample Pairs 2026-10-06
Humanoid Horizon: Extending Task Horizon in Whole-Body Loco-Manipulation via Parallel Training, Dynamic Starting, and Reward Gating 2026-10-06
Videos and results are available at https://haozhu...

Videos and results are available at https://haozhuo-zhang.github.io/Humanoid-Horizon-project-page/

zkLLMPoT: Efficient Zero Knowledge Proof of Training for Large Language Models 2026-10-06
Submitted to ICLR 2027...

Submitted to ICLR 2027

Quadratic Direct Forecast for Training Multi-Step Time-Series Forecast Models 2026-10-06
Accepted by ICLR 2026...

Accepted by ICLR 2026

LUMOS: Tracing Parametric Knowledge from Training Data to Behavioral Outputs in LLMs 2026-10-06
Accepted to NeurIPS 2026 (Poster)...

Accepted to NeurIPS 2026 (Poster)

Align, Then Correct: Training-Free Two-Stage Low-Rank Compensation for Extremely Quantized Large Language Models 2026-10-06 17 pages, 5 figures
Beyond Training from Scratch: Foundation Models for Data-Efficient and Generalizable Cardiac MRI Reconstruction 2026-10-06
Accepted at ECCVW 2026...

Accepted at ECCVW 2026

Enhancing Diffusion Language Models with Autoregressive Post-Training Weights 2026-10-06
Language Carries the Expert's Impression: Instrument-Anchored LLM Judges Transfer Counseling-Quality Assessment and Beat In-Domain Training 2026-10-06
Preprint. 25 pages, 2 figures...

Preprint. 25 pages, 2 figures

LionMuon: Alternating Spectral and Sign Descent for Efficient Training 2026-10-06
I mistakenly created a new submission instead of e...

I mistakenly created a new submission instead of editing the existing one

TF-PRVR: Training-Free Partially Relevant Video Retrieval 2026-10-06
Where Do Test-Time Scaling and Training Fall Short in Individual Stance Prediction? 2026-10-06
Rethinking Training Targets, Architectures and Data Quality for Universal Speech Enhancement 2026-10-06
Accepted at NeurIPS 2026...

Accepted at NeurIPS 2026

Inference

Title Date Comment
Prediction-powered inference for time series across space 2026-10-06
Accepted to TS-LIMITS Workshop at NeurIPS 2026...

Accepted to TS-LIMITS Workshop at NeurIPS 2026

Probabilistic Counterfactual Inference for Discrete Outcomes in Gaussian-Process Causal Models 2026-10-06
A Framework for Accelerating Transformer Inference on RISC-V for Edge AI 2026-10-06
Accepted for publication at the 2026 IEEE 33rd Int...

Accepted for publication at the 2026 IEEE 33rd International Conference on Electronics, Circuits and Systems (ICECS)

BitNest: Bit-Nested Speculative Decoding for Memory-Efficient LLM Inference Acceleration 2026-10-06
ActTune: Action-Aware Precision and GPU Operating-Point Adaptation for Energy-Efficient Vision-Language-Action Inference 2026-10-06
A computationally-tractable measure of global sensitivity for sampling-based Bayesian inference 2026-10-06
HintKD: Hint Knowledge Distillation for Bandwidth-Constrained Cloud-Edge Inference 2026-10-06
High-Dimensional Statistical Inference for Sparse Support Vector Machines 2026-10-06 7 figures
Enhancing LLMs with Cognitive-Affective Personality Inference for Simulating Human Social-Psychological Behavior 2026-10-06
Accepted at NeurIPS 2026...

Accepted at NeurIPS 2026

Where Do Two Populations of Persistence Diagrams Differ? Calibrated Local Inference at a Fixed Budget 2026-10-06
45 pages, 9 figures. Appendices with proofs and ad...

45 pages, 9 figures. Appendices with proofs and additional experiments. Under review

DySCo: Dynamic Sharding for Collaborative Edge-Cloud LLM Inference with Depth-Synchronized Batching 2026-10-06
article under submission...

article under submission

Robust Inference for Dyadic Data with Spatially Dependent Nodes 2026-10-06
SpliTEE: Fast and Private LLM Inference by Coupling GPU-Assisted Trusted Execution Environments with Differential Privacy 2026-10-06
ProximalFM: Amortized Proximal Causal Inference under Hidden Confounding 2026-10-06
Spectra: Exact Component Transport for Test-Time Prior Adaptation in Simulation-Based Inference 2026-10-06 41 pages, 7 figures
Flow-Transformed Implicit Processes for Function-Space Variational Inference 2026-10-06
27 pages, 5 figures, 11 tables. Accepted at the 40...

27 pages, 5 figures, 11 tables. Accepted at the 40th Conference on Neural Information Processing Systems (NeurIPS 2026)

Illusory Pattern Perception Drives Spurious Inference in Large Language Models 2026-10-06
accepted by NeurIPS 2026...

accepted by NeurIPS 2026

Persistent Memory in Multi-Agent LLM Inference: What It Costs, What It Buys, and When You Can Tell 2026-10-06
13 pages, 1 figure. Accepted as a poster at the Ma...

13 pages, 1 figure. Accepted as a poster at the Machine Learning for Systems Workshop, NeurIPS 2026

Readout Stability in Prefill-Only Decision Models:Zero-Label Prediction and Inference-Time Compute Allocation 2026-10-06
Semantic Split Inference for Remote Modulation Recognition 2026-10-06

Serving

Title Date Comment
Lachesis: Lifetime-Aware KV Cache Placement for Agent Serving across HBM and High-Bandwidth Flash 2026-10-06
ECO: Energy-Oriented Configuration Optimization for Attention FFN Disaggregated LLM Serving 2026-10-06
ServeLearnBench: How Well Can Agents Self-Improve from Serving Experience? 2026-10-06
28 pages, 13 figures. Code: https://github.com/Inf...

28 pages, 13 figures. Code: https://github.com/Infini-AI-Lab/ServeLearnBench. Project website: https://infini-ai-lab.github.io/ServeLearnBench

A Shape-Adaptive Architecture with Disaggregated Quantization for Efficient LLM Serving 2026-10-05
14 pages, 14 figures, 4 tables...

14 pages, 14 figures, 4 tables

Can Agent Harnesses and Inference Engines Hear Each Other? The HEAR Protocol for Agentic LLM Serving 2026-10-05
Jiaqi Zhao, Haodong Chen, and Jitai Hao contribute...

Jiaqi Zhao, Haodong Chen, and Jitai Hao contributed equally

Characterizing High Bandwidth Flash for LLM Serving 2026-10-05
MOLT: A Fine-Grained GPU Memory Sharing System for LLM Serving with Opportunistic Fine-Tuning 2026-10-05
Same Pieces, Different Servers: A Tetris Benchmark for AI Agents as Served 2026-10-05
11 pages, 7 figures, 3 tables. Code, raw logs and ...

11 pages, 7 figures, 3 tables. Code, raw logs and animations: https://github.com/KevinChunye/Diffusion-Tetris

Who Is Your Agent Serving? Provider-Side Indirect Prompt Injection in Proactive Agents 2026-10-04
30 pages, 7 figures, 13 tables...

30 pages, 7 figures, 13 tables

Beyond LLM Serving: Characterizing Vision-Language-Action Workloads for Embodied AI System Design 2026-10-04
To appear in the Proceedings of the 32nd ACM Inter...

To appear in the Proceedings of the 32nd ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS '27). 17 pages, 14 figures

Adaptive KV Cache Reuse for Fast Long-Context LLM Serving 2026-10-04
14 pages, Machine Learning...

14 pages, Machine Learning

RR-Evict: Fine-Grained Prefix Cache Eviction beyond LRU for Agentic LLM Serving 2026-10-02
Adaptive Co-Serving LLM Watermarking on Modern Inference Engines 2026-10-02
FluidPD: In-Place Elasticity for SLO-Aware Prefill-Decode Disaggregated LLM Serving 2026-10-02 13 pages, 11 figures
Coda: Exploiting Admission Flexibility for Coding-Agent Serving 2026-10-02
ePACT: Energy-Performance-Aware Commitment Tracking for LLM Serving 2026-10-01
Serving a Revisable World: Versioned Execution for Interruptible Agents 2026-10-01
Stated, Realized, and Optimal Aiming Strategies for the Tennis Serve: Experimental Evidence from Collegiate Athletes 2026-09-30 22 pages, 11 figures
Denoising Surface: Modeling and Predicting Inference Cost for Diffusion LLM Serving 2026-09-30
Taming Speculative Search for Test-Time Scaling in LLM Serving 2026-09-30 14 pages

Quantization

Title Date Comment
Test-Time Adaptation of Quantized ViTs via Single-Pass Quantizer-Aligned Recalibration 2026-10-06
44 pages, 6 figures. Code at https://github.com/ch...

44 pages, 6 figures. Code at https://github.com/chahh9808/QuAR

OSFP4: Joint Optimization of Diagonal Smoothing and Block Scales for NVFP4 Quantization 2026-10-06
Align, Then Correct: Training-Free Two-Stage Low-Rank Compensation for Extremely Quantized Large Language Models 2026-10-06 17 pages, 5 figures
ApexQuant: Data-Free Elastic Quantization by Residual Re-Isotropization 2026-10-06 25 pages, 6 figures
Tree-VQ: Progressive Image Compression from Pretrained Vector Quantizers 2026-10-06
Quantization Effects on Tool-Failure Recovery Vary Across Prompts and Evaluation Designs 2026-10-06
Accepted at the NeurIPS 2026 Workshop on Small Lan...

Accepted at the NeurIPS 2026 Workshop on Small Language Models for Agentic Systems (SLM-Agents). 7 pages, 2 figures, 2 tables, plus appendix

TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models 2026-10-06
Text Scores Do Not Establish Performance on Lexically Non-Diagnostic Speech Tasks: A Qwen2-Audio Quantization Case Study 2026-10-06
Activation Denoising: A Robustness View on Parallel vs Sequential LLM Quantization 2026-10-05
AlignQuant: Tile-Aligned Mixed-Precision Quantization for Efficient LLM Generation 2026-10-05
A Shape-Adaptive Architecture with Disaggregated Quantization for Efficient LLM Serving 2026-10-05
14 pages, 14 figures, 4 tables...

14 pages, 14 figures, 4 tables

QUASAR: Lowering the Loss Floor of Quantization-Aware Training with Loss-Aware Reconstruction 2026-10-05 44 pages
Exact Unlearning via Quantized Sufficient Statistics 2026-10-05
33 pages, 20 figures, 15 tables. Accepted at NeurI...

33 pages, 20 figures, 15 tables. Accepted at NeurIPS 2026

SoloQ: Calibration-Free Quantization for Diffusion Language Models 2026-10-05
CentriQ: Calibration-Free Quantization of Diffusion Transformers via Exact Mean Centering 2026-10-05
Code and project page will be released soon...

Code and project page will be released soon

Shared Stopping Decisions Change Answers in HQQ Cache Quantization 2026-10-05
Anosognosia in LLMs: Probing Self-Awareness of Quantized Computational Substrate 2026-10-05
9 main pages with appendix...

9 main pages with appendix

Which Decisions Low-Bit Quantization Breaks, and How to Predict Them 2026-10-05
37 pages, 9 figures, 12 tables. Preprint, under re...

37 pages, 9 figures, 12 tables. Preprint, under review

Differentiable Bit-Widths: Co-optimizing Pruning and Quantization via SVD for Ultra-Efficient LLM Compression 2026-10-05
Accepted to Advances in Neural Information Process...

Accepted to Advances in Neural Information Processing Systems (NeurIPS 2026)

Dissecting Quantization Error: A Concentration-Alignment Perspective 2026-10-05

RL

Title Date Comment
QF3: Fast Flow RL with Filtered Q-Gradients 2026-10-06
Project page: https://qf3-rl.github.io/...

Project page: https://qf3-rl.github.io/

Selective Transfer of RL Updates for Visual Reasoning 2026-10-06
NeMo-DCR: Bit-Exact Delta-Compressed Refit for Scalable Agentic RL at Trillion-Parameter Scale 2026-10-06
The code is open-sourced in NVIDIA NeMo RL PR #244...

The code is open-sourced in NVIDIA NeMo RL PR #2444 at NVIDIA-NeMo/RL#2444

FC-SWE: Failure-Conditioned RL for Long-Horizon Software Engineering Agents 2026-10-06
23 pages, 8 figures, 6 tables...

23 pages, 8 figures, 6 tables

EvoHarness-RL: Learning Runtime Harness Coordination for Self-Evolving Agents 2026-10-05
Accepted to LLA@COLM 2026...

Accepted to LLA@COLM 2026

NFTR: From Provable Mode-Averaging to Geodesic Subgoal Selection in Offline Goal-Conditioned RL 2026-10-05
Do Small Language Models Learn to Negotiate? A Controlled Scaling Study of RL-Trained Sellers 2026-10-05
20 pages, 3 figures, 9 tables. Accepted (poster) a...

20 pages, 3 figures, 9 tables. Accepted (poster) at the NeurIPS 2026 Workshop on SLMs for Agentic Systems (SLM-Agents), Paris

PlatoLTL: Scaling LTL-Guided Multi-Task RL 2026-10-05
16 pages, 2 figures (main paper). 29 pages, 8 figu...

16 pages, 2 figures (main paper). 29 pages, 8 figures (appendix)

Transfer-Stratified On-Policy Distillation for RL-Improved Reasoning Teachers 2026-10-05
HuatuoGPT-3: RL-Only Domain Adaptation from Base Models 2026-10-05
Extended version of "OnePO: Direct One-stage Polic...

Extended version of "OnePO: Direct One-stage Policy Optimization for SFT-free Domain Adaptation", accepted at ICML 2026, with additional analysis and scaling to HuatuoGPT-3

Timestep Weighting: A Hidden Key to Effective ELBO-Based Flow-Matching RL 2026-10-05
10 pages for the main body...

10 pages for the main body

1000 Layer Networks for Self-Supervised RL: Scaling Depth Can Enable New Goal-Reaching Capabilities 2026-10-05
Published at NeurIPS 2025, Best Paper Award. Link ...

Published at NeurIPS 2025, Best Paper Award. Link to project website: https://wang-kevin3290.github.io/scaling-crl/

Divide-and-Conquer CoT: RL for Reducing Latency via Parallel Reasoning 2026-10-05
59 pages, 16 figures. v2: Added appendices on wall...

59 pages, 16 figures. v2: Added appendices on wall-clock latency, RL runs with fewer stages, and qualitative analysis of rollouts. Corrected the Olympiad Bench result for DSR-12K/Maj-DSR-3 (v1 omitted 5 out of 674 problems during evaluation)

An LLM-in-the-loop RL Framework for Bioinformatics Feature Selection 2026-10-04
How Should Teachers Be Prepared? RL on Student-Induced States for On-Policy Distillation 2026-10-04
CORE-RL: Confidence-Oriented Reliability Evaluation of Black-Box Reinforcement Learning Policies 2026-10-03
TrustMed-RL: Long-Horizon Reinforcement Learning for Evidence-Grounded Clinical Diagnosis 2026-10-03
How RL Reshapes LLM Reasoning: Transferability, Coverage, and Scaling Laws 2026-10-03
VenusRL: A Fully Disaggregated Agentic RL System with Priority Scheduling and Scalable Interaction 2026-10-02 18 pages, 19 figures
Coverage You Can Steer: Online Conformal Calibration for RL-Driven Hardware-Aware NAS 2026-10-02

Diffusion

Title Date Comment
PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation 2026-10-06
9 pages, 4 figures. Project website: https://song-...

9 pages, 4 figures. Project website: https://song-kun.github.io/pears

Rapid Fredholm stabilization of the Kuramoto--Sivashinsky equation with unrestricted, spatially-varying anti-diffusion 2026-10-06 46 pages
Local Epistemic Uncertainty Guided Active Sampling for Plug-and-play Diffusive Image Restoration 2026-10-06
12 Pages, 7 Figures, 5 Tables. Accepted to ACM Mul...

12 Pages, 7 Figures, 5 Tables. Accepted to ACM Multimedia 2026 Oral!

ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks 2026-10-06
Denoising Hierarchical Representations: Joint Continuous Diffusion for Language Modeling 2026-10-06 27 pages, 10 figures
Local Content-Style Control for Diffusion-based Image Stylization 2026-10-06
SIGGRAPH Asia 2026 Technical Communications. 4 pag...

SIGGRAPH Asia 2026 Technical Communications. 4 pages, 4 figures, 1 table. Supplemental material included as an ancillary file

Steering Diffusion Models to Rare Events with Sequential Monte Carlo 2026-10-06
A previous version of this work was presented at t...

A previous version of this work was presented at the NeurIPS 2026: AI for Stochastic Dynamics workshop

Feature Information Dynamics in Diffusion 2026-10-06
Accepted as poster at NeurIPS 2026. 28 pages, incl...

Accepted as poster at NeurIPS 2026. 28 pages, including references, appendices, and checklist

LiDAR Resolution Recovery via Foundation-Model-Guided Diffusion 2026-10-06
One for All, All for One: Coordinated Multi-Agent Diffusion Steering via Stochastic Optimal Control 2026-10-06
Enhancing Diffusion Language Models with Autoregressive Post-Training Weights 2026-10-06
D3S2: Diffusion-Guided Dataset Distillation for Semantic Segmentation 2026-10-06
Diffusion Model-Based Video Editing: A Survey 2026-10-06
24 pages, 16 figures, a project related to this pa...

24 pages, 16 figures, a project related to this paper can be found at https://github.com/wenhao728/awesome-diffusion-v2v

Learned Adaptive Multiresolution Diffusion Imaging 2026-10-06
Best-of-$N$ Guidance for Test-time Diffusion Alignment 2026-10-06 NeurIPS 2026
Physics-Informed Conditional Diffusion for Motion-Robust Retinal Temporal Laser Speckle Contrast Imaging 2026-10-06
Feature Space Analysis by Guided Diffusion Model 2026-10-06
Accepted to ACCV 2026, 27 pages, 13 figures, 1 tab...

Accepted to ACCV 2026, 27 pages, 13 figures, 1 table, codes: https://github.com/ccilab-doshisha/FeatDec

FedGuide: Diffusion Prior Alignment and Value Baseline Guidance for Heterogeneous Federated Reinforcement Learning 2026-10-06
Accepted to the Conference on Robot Learning (CoRL...

Accepted to the Conference on Robot Learning (CoRL), 2026. Spotlight presentation

Model-Based Diffusion Optimal Control for Multi-Robot Motion Planning 2026-10-06
Published in Robotics: Science and Systems (RSS), ...

Published in Robotics: Science and Systems (RSS), 2026

SpecFold: Folding Multi-Branch Redundancy for Faster Speculative Decoding in Diffusion Language Models 2026-10-06

DIT

Title Date Comment
DAGS: Disentangled Appearance-and-Geometry Steering of a Frozen Image DiT for Temporally Stabilized Generative Rendering 2026-10-01
5 pages, 3 figures, 2 tables. Accepted to SIGGRAPH...

5 pages, 3 figures, 2 tables. Accepted to SIGGRAPH Asia 2026 Technical Communications

TLC-DiT: Task-Aligned Local Visual Conditioning for Robust Multitask Robot Manipulation 2026-09-28 6 pages, 5 figures
FIS-DiT: Breaking the Few-Step Video Inference Barrier via Training-Free Frame Interleaved Sparsity 2026-09-26
RefAdapt-DiT: Adaptive Joint Attention for Reference-Conditioned Diffusion Transformers 2026-09-26
PixelFlow: Token-Level Workload Management for Efficient Distributed DiT Serving 2026-09-17
DiT-Garment: Garment Dynamics with Diffusion Transformers 2026-09-16
SPARK: Input-Conditioned Sparse Activation Modulation for Frozen DiT-based Super-Resolution 2026-09-03
Shiva-DiT: Residual-Based Differentiable Top-$k$ Selection for Efficient Diffusion Transformers 2026-09-02 37 pages
V-Link: Recovering Lost Visual Representations in Action DiT for Vision-Language-Action Models 2026-08-26
Beyond Text Conditioning: A Systematic Study of MLLM-DiT Fusion for Video Generation 2026-08-14
Now You Have My Healthy Attention: A U-DiT for Brain-MRI Inpainting 2026-08-09
Fluid-DiT: Graph-Free Diffusion Transformers for Fluid Flow Simulations Learning 2026-08-07
ED-DiT: Physics-Guided Diffusion Pretraining for Transferable Molecular Representations from Electron Density 2026-08-04
16 pages, 9 figures, 7 tables, including supplemen...

16 pages, 9 figures, 7 tables, including supplementary material

Zellige: Moldable Sequence Placement for Mixed Image-Video DiT Training 2026-08-02
I2VShield: An Efficient Proactive Defense Framework against DiT-based Image-to-Video Models 2026-07-29
Audio-Anchored Fusion of Multi-Ratio DiT Reconstruction Residuals for Cross-Domain Audio Deepfake Detection 2026-07-29
10 pages, 5 figures. Submitted to speech security ...

10 pages, 5 figures. Submitted to speech security conference. This work proposes a cross-domain audio deepfake detection framework based on bona-fide trained DiT multi-ratio reconstruction residuals and audio-anchored additive fusion, evaluated on ASVspoof 5 and real-world ITW datasets

MegaSlide-DiT: Memory-Centric Adaptation and Deformable Local Attention for Efficient Video Diffusion 2026-07-17
When Geometry Aligns: Dihedral Hidden-State Transformations in UNet, ViT, and DiT Architectures 2026-07-03
Accepted at the Conference on Lifelong Learning Ag...

Accepted at the Conference on Lifelong Learning Agents (CoLLAs), 2026

GF-DiT: Scheduling Parallelism for Diffusion Transformer Serving 2026-07-02
GMO-E$^2$DIT: Grounded Multi-Operation Editing for E-Commerce Images 2026-07-02

MOE

Title Date Comment
MASKerade: Token-Routed Mask Experts for Dense-to-MoE Upcycling 2026-10-06
Reading, Not Manipulating: Leveraging Router Logits for Multimodal Safety in MoE Vision-Language Models 2026-10-06
15 pages, 4 tables, 11 figures...

15 pages, 4 tables, 11 figures

TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models 2026-10-06
Memory-Efficient Expert Routing for Distributed MoE Training 2026-10-05
Structuring MoE Expert Selection for Agentic Reinforcement Learning 2026-10-05
Cascadia: Resident 975B MoE Inference on Eleven AI PCs 2026-10-05
BRANCH-MoE: Balance-Aware Tree Routing for Large Embedding Models 2026-10-05
26 pages, 7 tables, 2 figures...

26 pages, 7 tables, 2 figures

UniAE-MoE: A Unified Audio Encoder via Mixture of Experts 2026-10-05
CIPHER-MoE: Balancing Efficiency and Routing Fidelity in Trillion-Scale MoE Training 2026-10-05
Context-Tower Conversion Preserves Generation While Freezing Retains Knowledge: Low-Budget AR-to-Diffusion Conversion of MoE LLMs 2026-10-05
v2: author order updated...

v2: author order updated

HiNa-MoE: High-Performance, Non-Intrusive MoE Inference on CPUs with Matrix Engines 2026-10-04
Accepted by PACT 2026...

Accepted by PACT 2026

Cross-Layer Analysis of Thermal Tuning Stalls in Wafer-Scale Optical Interconnects for LLM MoE Training 2026-10-03
BigMoMo: Efficient Inference of Large-Scale MoE with Speculative Decoding on Mobile Devices 2026-10-03
14 pages,15 figures,preprint,under review...

14 pages,15 figures,preprint,under review

MaskCoFT: Masked Co-Adaptive Fine-Tuning for Memory-Efficient MoE Inference 2026-10-02
SPEAR: A Spectral-Disentangled MoE Neural Operator with Knowledge-Guided Expert Aggregation for Large-Scale PDE Pretraining 2026-10-02
Cross-Lingual Alignment for Decoder-Only Models using MoE Routers 2026-10-02
Frequency Is Not Sensitivity Identifying Safety-Sensitive Experts in Sparse MoE LLM 2026-10-02
GenomeOcean Anywhere: Private WebGPU Inference for Genome MoEs 2026-10-01
Withdrawn by the authors pending an institutional ...

Withdrawn by the authors pending an institutional intellectual property review

MoE-CORE: Coordinated Expert Offloading and Residency for Memory-Constrained MoE Inference 2026-10-01
ITC-MoE: Importance-guided Token-aware Compression for MoE Diffusion Language Models 2026-10-01

VLA

Title Date Comment
WareFly-VLA: A Vision-Language-Action Framework for UAV Navigation and Human Tracking in Smart Warehouses 2026-10-06
41 pages, 35 figures, 11 tables...

41 pages, 35 figures, 11 tables

MIM-VLA: Learning Physical Interaction Representations from Gripper Motor Feedback 2026-10-06
9 pages, 7 figures. Jaeyoung Lee and Jiyeon Koo co...

9 pages, 7 figures. Jaeyoung Lee and Jiyeon Koo contributed equally

VLA-ACL: Action-Consistent Visual Token Pruning for Efficient Vision-Language-Action Models 2026-10-06
GeoAlign: Beyond Semantics with State-Guided Spatial Alignment in VLA Models 2026-10-06
20 pages, 9 figures, 8 tables, including appendix...

20 pages, 9 figures, 8 tables, including appendix

Seeing the Invisible: Physics-Guided Visual Prompting for Temperature- and Radiation-Aware VLA Navigation 2026-10-06
8 pages, 7 figures, 2 tables...

8 pages, 7 figures, 2 tables

Learning from Hindsight for VLA Reinforcement Learning 2026-10-05
VLA-ZO: Fast Zeroth-Order Adaptation for Vision-Language-Action Models 2026-10-05
Do VLAs Understand and Adapt to the Objects They Handle, or Simply Replay Learned Behaviors? 2026-10-05 8 pages
Look Before You Leap: Distilling Tree Search into Action Evaluation for Frozen VLA Models 2026-10-05
How (and How Not) to Use Data Augmentation in VLA Post-Training 2026-10-05
Accepted at the NeurIPS 2026 workshop RoboPAD. Cod...

Accepted at the NeurIPS 2026 workshop RoboPAD. Code and videos at https://bramgrooten.nl/vla-augm/

OGAM: Connecting Systematic Testing to Runtime Assurance through Object-Grounded Attention Monitoring for VLA Policies 2026-10-05
What the Guard Misses, the Robot Executes: Implied Harm in VLA Instructions 2026-10-05
Under review at the SPAIS 2026 workshop (CoRL 2026...

Under review at the SPAIS 2026 workshop (CoRL 2026)

Beyond In-Distribution Preservation: Recovering Generalization in Quantized VLAs via Vulnerability-Oriented Tuning 2026-10-05
EvoMem-VLA: State-Evolution Memory for Long-Horizon Robot Manipulation 2026-10-04
DyQ-VLA: Temporal-Dynamic-Aware Quantization for Embodied Vision-Language-Action Models 2026-10-04
When and What to Prune? Stage-Aware Visual Token Pruning for Efficient VLA 2026-10-04
Accepted at NeurIPS 2026...

Accepted at NeurIPS 2026

GeoBridge-VLA: Geometry-Aware Residual Adaptation for Vision-Language-Action Models 2026-10-04
Triggering Generalist Reasoning via Predictive Uncertainty for Dual-System VLA 2026-10-04
Accepted by NeurIPS 2026...

Accepted by NeurIPS 2026

DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling 2026-10-04
PermVLA: Factorization Order as a Regularizer for VLA Learning 2026-10-03

World Model

Title Date Comment
World Models' Last Exam in Physics 2026-10-06
DepthWorld: 3D World Model for Robot Manipulation 2026-10-06
Accepted at the Conference on Robot Learning (CoRL...

Accepted at the Conference on Robot Learning (CoRL) 2026. Project page: https://www.jaibardhan.com/depthworld. 32 pages including supplementary material, 15 figures, 7 tables

CtrlCache: Accelerating Interactive Video World Models with Control-Aware Caching 2026-10-06
18 pages. Project page: https://wrecklong.github.i...

18 pages. Project page: https://wrecklong.github.io/CtrlCache/

AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model 2026-10-06
Code at https://github.com/Sarim-MBZUAI/advsim2rea...

Code at https://github.com/Sarim-MBZUAI/advsim2real

WorldSonus: Bringing Sound to Worlds 2026-10-06
25 pages, 4 figures, 16 tables. Project page: http...

25 pages, 4 figures, 16 tables. Project page: https://noizai.github.io/WorldSonus/

TAPDreamer: Transferable Adversarial Patches for World Action Models 2026-10-06
Project Page: https://tapdreamer.github.io...

Project Page: https://tapdreamer.github.io

RIWANav: Recursive World-Action Models with Self-Improvement for Urban Navigation 2026-10-06
Parallel Predictive World Models for Accurate and Efficient Long-Horizon Planning 2026-10-06
WareFly-VLA: A Vision-Language-Action Framework for UAV Navigation and Human Tracking in Smart Warehouses 2026-10-06
41 pages, 35 figures, 11 tables...

41 pages, 35 figures, 11 tables

A Belief-State World Model for Catheter Navigation under Sparse Fluoroscopy: A Planar Proof of Concept 2026-10-06
Presented at the IEEE/RSJ International Conference...

Presented at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026) Workshop on Surgical Digital Twins (SurgTwin). 4 pages, 2 figures

Federated Bayesian Surveillance of Mechanical Thrombectomy Adverse Events: A Population Risk Layer for Surgical Digital Twins 2026-10-06
Presented at the IEEE/RSJ International Conference...

Presented at the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2026) Workshop on Surgical Digital Twins (SurgTwin). 3 pages, 1 figure

How Much Planning Is Enough? Reducing Search and Computation in World-Model Planning 2026-10-06
Attention from Above: A Multimodal Model for Drone-Based Object Localization 2026-10-06
Published in the International Journal of Interact...

Published in the International Journal of Interactive Mobile Technologies

Learning in Dreams, Winning in Reality: A Continuous Dyna Loop for a Ten-Hero MOBA 2026-10-06
15 pages, 11 figures. Code, policies, evaluation a...

15 pages, 11 figures. Code, policies, evaluation and videos: https://github.com/JordyKieto/puffermoba-dyna

Commit While Futures Agree: Consequence-Aware Adaptive Action Chunking for Robot Manipulation 2026-10-06
18 pages, 10 figures, including appendix...

18 pages, 10 figures, including appendix

Adaptive Latent Capacity for World Models 2026-10-06
World Properties without World Models: Distributional Associations and the Interpretation of Decoding Results from Language Models 2026-10-06
22 pages, 3 figures, 10 tables. Substantially revi...

22 pages, 3 figures, 10 tables. Substantially revised to include analyses of full released Gurnee & Tegmark datasets with Llama-2 and Pythia comparisons; replaces the earlier 100-city, 194-figure analysis; also includes entity-level vectors, and pain and emotion decoding

Reperesentation Geometry Matters for Planning with JEPA World Models 2026-10-06 19 pages, 3 figures
Independent Multi-Agent Reinforcement Learning with Counterfactual Semantic-Social World Models 2026-10-06
Completion Aware Guidance for World Action Models 2026-10-06

video generation

Title Date Comment
World Models' Last Exam in Physics 2026-10-06
Backend-Agnostic Sparse Attention for Fast High-Resolution Visual Generation 2026-10-06
Dataset Biases and Shortcut Learning in Motion-Based AI-Generated Video Detection 2026-10-06
Transferable Spatial Temporal Coherence Adversarial Attack on Black-Box Vision Language Models for Autonomous Driving 2026-10-06
Compact Robot Policies Need Fine-Grained Visual Representations 2026-10-06
35 pages, 21 figures, 8 tables...

35 pages, 21 figures, 8 tables

No Corners Cut: State-Grounded Transitions for Mid-Stream Prompt Switches in Video Generation 2026-10-06
Page: https://anonymous.4open.science/w/No-Corners...

Page: https://anonymous.4open.science/w/No-Corners-Cut-6C5D/

Rethinking Visual Provenance: Detection and Watermarking Across Direct Visual Generation and LLM-Driven Code Rendering 2026-10-06
46 pages, 6 figures, 4 tables. Conceptual research...

46 pages, 6 figures, 4 tables. Conceptual research agenda; no experiments. Video: https://youtu.be/14SMl0d_e48. Project page: https://zhenggao-30.github.io/Rethinking-Visual-Provenance/

ChronoWorld: Camera-Controlled Consistent 4D World Generation via Spatiotemporal Cues and Geometric Reflections 2026-10-06
DensiTok: Making Feed-Forward 3D Gaussian Splatting See More Views Than It Is Given 2026-10-06
CCDF: A Benchmark Dataset for Deepfake Detection in Real-World Surveillance Footage 2026-10-06
Diverse Motion Customization via Control-based Dynamic Optimization 2026-10-06 Preprint
S2PD: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation 2026-10-05
Project Page: https://jefequien.github.io/S2PD/...

Project Page: https://jefequien.github.io/S2PD/

VideoGen-Agent: Reinforcing Video Generation Agents 2026-10-05
RealtimeWAM: One-Step Asynchronous World Action Models 2026-10-05
The code and checkpoints are available at $\href{h...

The code and checkpoints are available at $\href{https://github.com/ModelTC/LightX2V/tree/main/examples/realtimewam}{\text{this https URL}}$

SimForcing: Distilling Simulation Motion Priors into Real-Domain Robot World Models 2026-10-05
Code: https://github.com/Wang-Xiaodong1899/SimForc...

Code: https://github.com/Wang-Xiaodong1899/SimForcing

Keepsake: Selective Spatial Memory for Long-Horizon Video Generation 2026-10-05
MTOR: Generalizable AI-Generated Video Detection with Multimodal Semantics and Temporal Over-Regularity 2026-10-05
18 pages, 4 figures, 19 tables...

18 pages, 4 figures, 19 tables

TasteRoute: Personalized Routing for Video Generation 2026-10-05
Custom Forcing: Training-Free Subject Customization for Autoregressive Video Generation 2026-10-05
31 pages. Project page: https://gustn9609.github.i...

31 pages. Project page: https://gustn9609.github.io/custom-forcing/

Level-of-Token Diffusion 2026-10-05

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentation

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions