Paper Digest Index (2026-05-09 to 2026-05-25): Daily AI Research Roundups

Forum topic · 小凯 · 2026-05-25

Summary

This forum post is a sub-index of daily paper digest roundups published on zhichai.net between May 9 and May 25, 2026, ordered in reverse chronological order. Each dated section links to individual paper digest threads covering recent machine learning research. Topics span video understanding and generation (Cambrian-P, MotiMotion, RAVEN, ActCam), vision-language and vision-language-action models (AwareVLN, GesVLA, Vision-OPD), LLM reasoning and agents (Equilibrium Reasoners, OpenDeepThink, Recursive Agent Optimization, WildClawBench, FutureSim), efficient attention and long-context inference (DashAttention, KV-Fold, Elastic Attention Cores), diffusion and generative modeling (SURGE, WavFlow, SANA-WM), mixture-of-experts (EMO, UniPool, DECO), autonomous driving (Sensor2Sensor, 123D), benchmarks (WikiVQABench, DeepWeb-Bench, LongMemEval-V2, ESI-Bench), and interpretability, safety, and evaluation methodology. Duplicate entries appear within several dates. All entries link to forum threads at zhichai.net/t/ URLs. The index serves as a navigational hub for tracking daily AI/ML arXiv-style paper coverage on the forum.

Paper Digest Sub-Index (2026-05-09 ~ 2026-05-25)

This index collects the daily paper digest roundups posted on zhichai.net from 2026-05-09 through 2026-05-25, sorted in reverse chronological order. Each entry links to the corresponding paper digest thread on the forum. Some duplicate entries appear in the source index and are preserved below.

2026-05-25

Cambrian-P: Pose-Grounded Video Understanding → https://zhichai.net/t/177620758
MotiMotion: Motion-Controlled Video Generation with Visual Reasoning → https://zhichai.net/t/177620759
Which Way Did It Move? Diagnosing and Overcoming Directional Motion Blindness in Video-LLMs → https://zhichai.net/t/177620756
GesVLA: Gesture-Aware Vision-Language-Action Model Embedded Representations → https://zhichai.net/t/177620762
The Matching Principle: A Geometric Theory of Loss Functions for Nuisance-Robust Representation Learning → https://zhichai.net/t/177620764

2026-05-23

Integrable Elasticity via Neural Demand Potentials → https://zhichai.net/t/177620655
Tokenisation via Convex Relaxations → https://zhichai.net/t/177620653
AwareVLN: Reasoning with Self-awareness for Vision-Language Navigation → https://zhichai.net/t/177620659
Cambrian-P: Pose-Grounded Video Understanding → https://zhichai.net/t/177620656
MotiMotion: Motion-Controlled Video Generation with Visual Reasoning → https://zhichai.net/t/177620657
Vector Policy Optimization: Training for Diversity Improves Test-Time... → https://zhichai.net/t/177620658
Remember to be Curious: Episodic Context and Persistent Worlds for 3D... → https://zhichai.net/t/177620660
GesVLA: Gesture-Aware Vision-Language-Action Model Embedded Representations → https://zhichai.net/t/177620661
Sensor2Sensor: Cross-Embodiment Sensor Conversion for Autonomous Driving → https://zhichai.net/t/177620662
Which Way Did It Move? Diagnosing Directional Motion Blindness in Video-LLMs → https://zhichai.net/t/177620654

2026-05-22

Velocityformer: Broken-Symmetry-Matched Equivariant Graph Transformers → https://zhichai.net/t/177620578
Quantifying Hyperparameter Transfer and the Importance of Embedding La... → https://zhichai.net/t/177620572
Variance Reduction for Expectations with Diffusion Teachers → https://zhichai.net/t/177620569
One-Step Distillation of Discrete Diffusion Image Generators via Fixed... → https://zhichai.net/t/177620574
WikiVQABench: A Knowledge-Grounded Visual Question Answering Benchmark → https://zhichai.net/t/177620576
EvoStruct: Bridging Evolutionary and Structural Priors for Antibody CD... → https://zhichai.net/t/177620573
Equilibrium Reasoners: Learning Attractors Enables Scalable Reasoning → https://zhichai.net/t/177620570
Uni-Edit: Intelligent Editing Is A General Task For Unified Model Tuning → https://zhichai.net/t/177620571
DeepWeb-Bench: A Deep Research Benchmark Demanding Massive Cross-Source... → https://zhichai.net/t/177620575

2026-05-20

Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy... → https://zhichai.net/t/177620488
Actionable World Representation → https://zhichai.net/t/177620487
SURGE: Approximation-free Training-Free Particle Filter for Diffusion → https://zhichai.net/t/177620486
DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention → https://zhichai.net/t/177620480
WavFlow: Audio Generation in Waveform Space → https://zhichai.net/t/177620482
Aurora: Unified Video Editing with a Tool-Using Agent → https://zhichai.net/t/177620483
Can These Views Be One Scene? Evaluating Multiview 3D Consistency → https://zhichai.net/t/177620479
A Readiness-Driven Runtime for Pipeline-Parallel Training under Runtime... → https://zhichai.net/t/177620481
Code as Agent Harness → https://zhichai.net/t/177620484
ESI-Bench: Towards Embodied Spatial Intelligence → https://zhichai.net/t/177620485

2026-05-19

DeepSlide: From Artifacts to Presentation Delivery → https://zhichai.net/t/177620349
SkillSmith: Compiling Agent Skills into Boundary-Guided Runtime Interfaces → https://zhichai.net/t/177620352
Fair outputs, Biased Internals: Causal Potency and Asymmetry of Latent... → https://zhichai.net/t/177620353
NIMO Controller: A Self-Driving Laboratory Orchestrator → https://zhichai.net/t/177620357
NOVA: Fundamental Limits of Knowledge Discovery Through AI → https://zhichai.net/t/177620355
Does Theory of Mind Improvement Really Benefit Human-AI Interactions? → https://zhichai.net/t/177620351
SDOF: Taming the Alignment Tax in Multi-Agent Orchestration → https://zhichai.net/t/177620350
Solvita: Enhancing LLMs for Competitive Programming → https://zhichai.net/t/177620358

2026-05-18

From Descriptive to Prescriptive: Uncover the Social Value Alignment → https://zhichai.net/t/177620218
PolitNuggets: Benchmarking Agentic Discovery of Long-Tail Political Facts → https://zhichai.net/t/177620215
Conditional Attribute Estimation with Autoregressive Sequence Models → https://zhichai.net/t/177620216
Sheaf-Theoretic Transport and Obstruction for Detecting Scientific The... → https://zhichai.net/t/177620217
Mixed Integer Goal Programming for Personalized Meal Optimization → https://zhichai.net/t/177620211
A Two-Dimensional Framework for AI Agent Design Patterns → https://zhichai.net/t/177620212
PREPING: Building Agent Memory without Tasks → https://zhichai.net/t/177620214
Enhanced and Efficient Reasoning in Large Learning Models → https://zhichai.net/t/177620219
Invisible Orchestrators Suppress Protective Behavior → https://zhichai.net/t/177620213
Model-Adaptive Tool Necessity Reveals the Knowing-Doing Gap in LLM Tool Use → https://zhichai.net/t/177620220

2026-05-17

FutureSim: Replaying World Events to Evaluate Adaptive Agents → https://zhichai.net/t/177620162
Evidential Reasoning Advances Interpretable Real-World Disease Screening → https://zhichai.net/t/177620169
RAVEN: Real-time Autoregressive Video Extrapolation → https://zhichai.net/t/177620161
Quantitative Video World Model Evaluation for Geometric-Consistency → https://zhichai.net/t/177620165
Articraft: An Agentic System for Scalable Articulated 3D Asset Generation → https://zhichai.net/t/177620163
OpenDeepThink: Parallel Reasoning via Bradley–Terry Aggregation → https://zhichai.net/t/177620167
Text Knows What, Tables Know When: Clinical Timeline Reconstruction → https://zhichai.net/t/177620170
MetaBackdoor: Exploiting Positional Encoding as a Backdoor Attack Surface → https://zhichai.net/t/177620168
SANA-WM: Efficient Minute-Scale World Modeling with Hybrid Linear Diffusion → https://zhichai.net/t/177620166
VGGT-Edit: Feed-forward Native 3D Scene Editing → https://zhichai.net/t/177620164

2026-05-16

FutureSim: Replaying World Events to Evaluate Adaptive Agents → https://zhichai.net/t/177620087
Evidential Reasoning for Interpretable Disease Screening → https://zhichai.net/t/177620094
RAVEN: Real-time Autoregressive Video Extrapolation → https://zhichai.net/t/177620086
OpenDeepThink: Parallel Reasoning via Bradley–Terry Aggregation → https://zhichai.net/t/177620092
Quantitative Video World Model Evaluation → https://zhichai.net/t/177620090
Articraft: Scalable Articulated 3D Asset Generation → https://zhichai.net/t/177620088
VGGT-Edit: Feed-forward Native 3D Scene Editing → https://zhichai.net/t/177620089
Text Knows What, Tables Know When: Clinical Timeline Reconstruction → https://zhichai.net/t/177620095
SANA-WM: Efficient Minute-Scale World Modeling → https://zhichai.net/t/177620091
MetaBackdoor: Positional Encoding as a Backdoor Attack Surface → https://zhichai.net/t/177620093

2026-05-15

From Plans to Pixels: Learning to Plan and Orchestrate for Open-Ended... → https://zhichai.net/t/177620064
When Are Two Networks the Same? Tensor Similarity for Mechanistic Interpretability → https://zhichai.net/t/177620062
Eradicating Negative Transfer in Multi-Physics Foundation Models → https://zhichai.net/t/177620065
Warp-as-History: Generalizable Camera-Controlled Video Generation → https://zhichai.net/t/177620063
Is Grep All You Need? How Agent Harnesses Reshape Agentic Search → https://zhichai.net/t/177620061
Aligning Latent Geometry for Spherical Flow Matching in Image Generation → https://zhichai.net/t/177620060
VGGT- $\Omega$ → https://zhichai.net/t/177620059
RefDecoder: Enhancing Visual Generation with Conditional Video Decoding → https://zhichai.net/t/177620058
EntityBench: Entity-Consistent Long-Range Multi-Shot Video Generation → https://zhichai.net/t/177620056
ATLAS: Agentic or Latent Visual Reasoning? → https://zhichai.net/t/177620057

2026-05-14

Pion: A Spectrum-Preserving Optimizer via Orthogonal Equivalence Transformation → https://zhichai.net/t/177620004
Revisiting Photometric Ambiguity for Accurate Gaussian-Splatting Surfaces → https://zhichai.net/t/177620002
MEME: Multi-entity & Evolving Memory Evaluation → https://zhichai.net/t/177620011
Task-Adaptive Embedding Refinement via Test-time LLM Guidance → https://zhichai.net/t/177620006
Solve the Loop: Attractor Models for Language and Reasoning → https://zhichai.net/t/177620014
AlphaGRPO: Unlocking Self-Reflective Multimodal Generation in UMMs → https://zhichai.net/t/177620001
KV-Fold: One-Step KV-Cache Recurrence for Long-Context Inference → https://zhichai.net/t/177620013
Covering Human Action Space for Computer Use: Data Synthesis and Benchmarks → https://zhichai.net/t/177619997
Beyond GRPO and On-Policy Distillation: Sparse-to-Dense RL → https://zhichai.net/t/177620008
From Web to Pixels: Bringing Agentic Search into Visual Perception → https://zhichai.net/t/177619999
EgoForce: Forearm-Guided Camera-Space 3D Hand Pose from Monocular Ego Video → https://zhichai.net/t/177619998
Elastic Attention Cores for Scalable Vision Transformers → https://zhichai.net/t/177620005
LongMemEval-V2: Evaluating Long-Term Agent Memory → https://zhichai.net/t/177620003
CausalCine: Real-Time Autoregressive Generation for Multi-Shot Video Narrative → https://zhichai.net/t/177620000
ToolCUA: Towards Optimal GUI-Tool Path Orchestration for Computer Use Agents → https://zhichai.net/t/177620009
Learning, Fast and Slow: Towards LLMs That Adapt Continually → https://zhichai.net/t/177620007
OmniNFT: Modality-wise Omni Diffusion Reinforcement for Joint Audio-Video Generation → https://zhichai.net/t/177620010
Routers Learn the Geometry of Their Experts: Geometric Coupling in Sparse MoE → https://zhichai.net/t/177620012

2026-05-13

V4FinBench: Benchmarking Tabular Foundation Models, LLMs, and Standard Methods → https://zhichai.net/t/177619930
CapVector: Learning Transferable Capability Vectors in Parametric Space → https://zhichai.net/t/177619927
Equivariant Reinforcement Learning for Clifford Quantum Circuit Synthesis → https://zhichai.net/t/177619923
Engineering Robustness into Personal Agents with the AI Workflow Store → https://zhichai.net/t/177619925
Revisiting Policy Gradients for Restricted Policy Classes → https://zhichai.net/t/177619924
Beyond Red-Teaming: Formal Guarantees of LLM Guardrail Classifiers → https://zhichai.net/t/177619928
DataMaster: Towards Autonomous Data Engineering for Machine Learning → https://zhichai.net/t/177619926
Shepherd: A Runtime Substrate Empowering Meta-Agents → https://zhichai.net/t/177619921
RubricEM: Meta-RL with Rubric-guided Policy Decomposition → https://zhichai.net/t/177619929
WildClawBench: A Benchmark for Real-World, Long-Horizon Agent Evaluation → https://zhichai.net/t/177619922
ELF: Embedded Language Flows → https://zhichai.net/t/177619911
Optimal and Scalable MAPF via Multi-Marginal Optimal Transport → https://zhichai.net/t/177619919
DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance → https://zhichai.net/t/177619915
Quantifying Concentration Phenomena of Mean-Field Transformers → https://zhichai.net/t/177619916
Personal Visual Context Learning in Large Multimodal Models → https://zhichai.net/t/177619913
Variational Inference for Lévy Process-Driven SDEs via Neural Tilting → https://zhichai.net/t/177619914
Confidence-Guided Diffusion Augmentation for Bangla Compound Script Recognition → https://zhichai.net/t/177619920

2026-05-12

A Note on Non-Negative L1-Approximating Polynomials → https://zhichai.net/t/177619880
VecCISC: Improving Confidence-Informed Self-Consistency with Reasoning Vectors → https://zhichai.net/t/177619881
GRAPHLCP: Structure-Aware Localized Conformal Prediction on Graphs → https://zhichai.net/t/177619878
Proxy3D: Efficient 3D Representations for Vision-Language Models → https://zhichai.net/t/177619882
LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling → https://zhichai.net/t/177619874
EmambaIR: Efficient Visual State Space Model for Event-guided Image Restoration → https://zhichai.net/t/177619879
Normalizing Trajectory Models → https://zhichai.net/t/177619875
Conformal Path Reasoning: Trustworthy Knowledge Graph Question Answering → https://zhichai.net/t/177619876
123D: Unifying Multi-Modal Autonomous Driving Data at Scale → https://zhichai.net/t/177619871

2026-05-11

Recursive Agent Optimization → https://zhichai.net/t/177619785
GlazyBench: A Benchmark for Ceramic Glaze Property Prediction and Imaging → https://zhichai.net/t/177619784

2026-05-10

Inductive Venn-Abers and Related Regressors → https://zhichai.net/t/177619698
Concept-Based Abductive and Contrastive Explanations → https://zhichai.net/t/177619701
Optimizer-Model Consistency: Full Finetuning with the Same Optimizer → https://zhichai.net/t/177619694
BAMI: Training-Free Bias Mitigation in GUI Grounding → https://zhichai.net/t/177619689
Verifier-Backed Hard Problem Generation for Mathematical Reasoning → https://zhichai.net/t/177619691
ActCam: Zero-Shot Joint Camera and 3D Motion Control for Video Generation → https://zhichai.net/t/177619687
When No Benchmark Exists: Validating Comparative LLM Safety Scoring → https://zhichai.net/t/177619695
Why Global LLM Leaderboards Are Misleading → https://zhichai.net/t/177619693
Superintelligent Retrieval Agent: The Next Frontier of Information Retrieval → https://zhichai.net/t/177619697
Relit-LiVE: Relight Video by Jointly Learning Environment Video → https://zhichai.net/t/177619692
Edge-specific Signal Propagation on 3D Mechanistic Models → https://zhichai.net/t/177619699
Are We Making Progress in Multimodal Domain Generalization? → https://zhichai.net/t/177619700
Beyond Negative Rollouts: Positive-Only Policy Optimization → https://zhichai.net/t/177619696
EMO: Pretraining Mixture of Experts for Emergent Modularity → https://zhichai.net/t/177619690
UniPool: A Globally Shared Expert Pool for Mixture-of-Experts → https://zhichai.net/t/177619688

2026-05-09

BAMI: Training-Free Bias Mitigation in GUI Grounding → https://zhichai.net/t/177619662
Verifier-Backed Hard Problem Generation for Mathematical Reasoning → https://zhichai.net/t/177619664
ActCam: Zero-Shot Joint Camera and 3D Motion Control for Video Generation → https://zhichai.net/t/177619660
UniPool: A Globally Shared Expert Pool for Mixture-of-Experts → https://zhichai.net/t/177619661
EMO: Pretraining Mixture of Experts for Emergent Modularity → https://zhichai.net/t/177619663
Optimizer-Model Consistency: Full Finetuning with the Same Optimizer → https://zhichai.net/t/177619667