INDEX
Every paper Pith has read and judged, in one searchable index.
-
cs.AI arXiv submitted 2026-08-03Answering before reading causes many agentic RAG errors
Daeyoung Roh +1 · “Before Reasoning Can Fail: Pre-Evidence Procedural Failures in Agentic RAG”
2608.02011 -
cs.AI arXiv submitted 2026-08-03Stop search agents when evidence covers every needed fact
Daeyoung Roh +1 · “HALT: Verification-Aware Stopping for Retrieval-Augmented Search Agents”
2608.02009 -
cs.CV arXiv submitted 2026-08-03Motion tracks beat pixel colors for unsynced 3D video
Junyu Zhu +5 · “ASTRA: Asynchronous Spatio-Temporal Reconstruction via Trajectory Alignment”
2608.02006 -
cs.AI arXiv submitted 2026-08-03Gated skill updates tether LLM agents to equilibrium
Zhaofeng Zhang +5 · “Evolving in the Agent Jungle via History-Informed Opponent Awareness”
2608.02005 -
cs.DL arXiv submitted 2026-08-03BIP! Ranker computes eight citation metrics on billion-edge graphs
Ilias Kanellos +2 · “BIP! Ranker: A Software Library for Citation-Based Impact Indicators on Large-Scale Graphs”
2608.02004 -
cs.SE arXiv submitted 2026-08-03Best coding agent finds under 23% of hard repository vulnerabilities
Kexing Ji +9 · “VulnGym: Benchmarking Coding Agents for Repository-Level Vulnerability Detection”
2608.02001 -
eess.SY arXiv submitted 2026-08-03Properly tuned sliding-mode observer beats I&I in replication
A.J. Munoz-Vazquez +1 · “A Fair Comparison of Sliding-Mode and Immersion-and-Invariance Observers”
2608.02000 -
cs.DC arXiv submitted 2026-08-03Lyapunov scheduler times VLM offloads for timely driving
Mengmeng Zhu +3 · “TALSC: Timeliness-Aware Large-Small VLM Collaboration for Infrastructure-Assisted Autonomous Driving”
2608.01998 -
cs.LG arXiv submitted 2026-08-03Optimizer switcher tops single-method training on 6 of 8 benchmarks
Alok Kumar Pandey +3 · “AOS: Adaptive Optimizer Switching via Training-State Signals for Faster Convergence and Better Generalization”
2608.01997 -
cs.AI arXiv submitted 2026-08-03Workflow, not model, drives long-horizon AI research behavior
Aon Safdar +1 · “Long-Horizon Autonomous Architecture Research with a Language-Model Agent: A Behavioural Case Study”
2608.01995 -
cs.CE arXiv submitted 2026-08-03Transport certificate shows switching diffusions contract exponentially
Yutong Zhu +1 · “Mode-Weighted Transport Certificates for State-Dependent Reflected Switching Diffusions”
2608.01992 -
cs.CV arXiv submitted 2026-08-03Zero-parameter affinity loss cuts ImageNet FID in diffusion transformers
Zong-Wei Hong +5 · “SPARE: Structural Parameter-Free Affinity Regularization for Flow Matching”
2608.01990 -
cs.CV arXiv submitted 2026-08-03AI detector names the fake pixels and explains why
Kun Guo +5 · “Grounding and Explaining Visual Evidence for AI-Generated Image Detection in Human-Centric Scenes”
2608.01988 -
cs.CC arXiv submitted 2026-08-03Full Euclidean transcript defeats constant-depth arithmetic
Amiel Ferman · “The Complete Extended Euclidean Scheme Is Not in Piecewise Arithmetic $\mathrm{AC}^0$”
2608.01987 -
cs.CE arXiv submitted 2026-08-03One RL policy matches reservoir history in 200 simulator calls
Nam-Phong Huu Nguyen +2 · “Physics-Constrained Soft Actor-Critic for Simulator-in-the-Loop Petroleum Reservoir History Matching”
2608.01986 -
cs.CV arXiv submitted 2026-08-03Token pruning via noise throttling keeps 96.5% accuracy
Landi He +3 · “DiffPrune: differentiable information throttling for token pruning in vision-language models”
2608.01985 -
cs.LO arXiv submitted 2026-08-03Parser checks drop from a day to minutes
Jort van Leenen +1 · “Octopus: Practical Equivalence Checking of P4 Packet Parsers”
2608.01982 -
cs.RO arXiv submitted 2026-08-03Cut painters' hand motion by letting the robot learn their comfort
C. Cella +4 · “Adaptive Human-Robot Collaborative Painting Combining Preference-Based Optimization and Dynamic Motion Primitives”
2608.01981 -
cs.CV arXiv submitted 2026-08-03Video model learns when to think
Jingqi Tian +9 · “AdaThinkV: Adaptive Thinking for Token-Efficient Video Reasoning”
2608.01980 -
cs.CV arXiv submitted 2026-08-03Per-question budgets beat fixed pruning in OCR at half tokens
Zizhong Ding +6 · “ET-Prune: Evidence-Aware Dynamic Budgeting for Visual Token Pruning in Text-Rich MLLMs”
2608.01979 -
cs.CV arXiv submitted 2026-08-03Single proxy avatar animates any portrait with emotion at 32 FPS
Haijie Yang +8 · “Proxy Avatar Meets Low-Rank Caching: Real-Time One-Shot Emotion-Controllable Portrait Animation”
2608.01978 -
cs.CV arXiv submitted 2026-08-03New benchmark shows models fail SVG spatial and structural checks
Yiming Wang +3 · “SVGEval: A Vision-Grounded Framework for Perceptual-Quality Benchmarking and Evaluation in Text-to-SVG Generation”
2608.01977 -
cs.SE arXiv submitted 2026-08-03Trace-and-log tool localizes LLM inference failures
Ruilin Xu +3 · “TELLER: Non-intrusive Cross-Layer Root-Cause Analysis for LLM Inference”
2608.01975 -
cs.HC arXiv submitted 2026-08-03Splatting looks real, but mesh wins VR tourism UX
Maximilian Warsinke +5 · “3D Gaussian Splatting and Mesh-Based Digital Twins: An Exploratory Study for Virtual Reality Tourism”
2608.01969 -
cs.LG arXiv submitted 2026-08-03A chaotic probe sorts four frozen transformers into known families
Kunal Kumar Pant +1 · “ChaosProbe: A Neurochaotic Lens on Frozen Transformer Input-Embedding Spaces”
2608.01968 -
cs.GT arXiv submitted 2026-08-03Augmented MS-potential equals CMOP potential when action counts match
Robert P. Gilles · “Second-Order Potentials for Finite Games: Existence, Characterisation, and Game Decomposition”
2608.01967 -
cs.LO arXiv submitted 2026-08-03Machine-checked STARK proof bounds false acceptance probability
Diego Marmsoler · “Isabelle/STARK: A Formalization of zk-STARK in Isabelle/HOL”
2608.01965 -
cs.CV arXiv submitted 2026-08-03Audited task state lifts long-horizon agents from 52% to 81%
Ziyu Ma +7 · “LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks”
2608.01964 -
cs.CV arXiv submitted 2026-08-03New Sentinel-1 ship dataset offers 55,759 labeled ships in float format
Horst Hammer +6 · “OSSDD - a New Open Dataset for Sentinel-1 Ship Detection”
2608.01963 -
cs.CV arXiv submitted 2026-08-03Fourier motion beats polynomials in dynamic 4D Gaussian splatting
Zhengyang Zhang +8 · “FAST-GS: Frequency Aware Space-time Gaussian Splatting for Photorealistic Dynamic Novel View Synthesis”
2608.01958 -
cs.ET arXiv submitted 2026-08-03Self-healing pipelines need no new tech
Solomon Eshun +6 · “Agentic Self-Healing for Data and AI Pipelines: An Affordable Vendor-Agnostic Architecture using Open-Source Software”
2608.01955 -
cs.RO arXiv submitted 2026-08-03Shiny parts: 56% better 3D reconstruction coverage
G. F. Preziosa +4 · “FRA-NBV: A Fast and Reflectivity-Aware Next-Best-View Strategy”
2608.01950 -
cs.IR arXiv submitted 2026-08-03Self-triggering push system boosts user active days 0.28%
Zhao-Yu Zhang +11 · “A Self-Triggered Agentic Push Recommendation System”
2608.01949 -
cs.CV arXiv submitted 2026-08-03Simulated event streams cover 106 hours of untrimmed action
Cheng-Yao Hong +5 · “Event ActivityNet: A Large-Scale Simulated-Event Benchmark for Untrimmed Action Understanding”
2608.01948 -
q-bio.NC arXiv submitted 2026-08-03Dividing neural drive by a gain signal preserves continuous memory
Zhaotian Gu +5 · “Divisive Normalization Shapes Low-Rank Slow Manifolds for Continuous Working Memory”
2608.01947 -
cs.AR arXiv submitted 2026-08-037.6x speedup for per-cycle power toggles
Wenkai Li +6 · “LEAP: A Self-Supervised Per-Cycle Toggle Propagation Model Supports Fast, Transferable, and Early Analysis of Layout Power”
2608.01946 -
cs.CE arXiv submitted 2026-08-03Neural surrogates reproduce induction-welding heat fields to 0.3–0.4 °C
Wouter J. Schuttert +2 · “A manifold-aware Neural ODE surrogate model for stochastic induction heating with anisotropic electrical conductivity”
2608.01945 -
cs.CV arXiv submitted 2026-08-03One visual proxy unifies human motion and camera control
Liming Tan +5 · “UniMoCa: Unifying Motion and Camera Controls as Visual Proxies for Faithful Human Video Generation”
2608.01944 -
cs.CV arXiv submitted 2026-08-031,000-prompt test finds culture blind spots in video AI
Xianjing Han +5 · “CultureVidBench: Benchmarking Cultural Understanding in Text-to-Video Generation”
2608.01942 -
cs.CL arXiv submitted 2026-08-03Ancient Greek vowel-length model hits 92% accuracy on gold benchmark
Albin Th\"orn Cleland +1 · “Automatic Annotation of Ancient Greek Vowel Length”
2608.01935 -
cs.CV arXiv submitted 2026-08-03Blocking stale reasoning lifts VLM visual updates from 35% to 54%
Wenxiao Fan +11 · “Recompute or Reuse? Diagnosing and Mitigating Textual Shortcuts in VLM Self-Reflection”
2608.01930 -
math.CO arXiv submitted 2026-08-03Minimal rewiring moves grow quadratically with graph max degree
Jane Ivy Coons +1 · “Markov and lattice bases for Forman-Ricci curvature of graphs”
2608.01929 -
eess.SY arXiv submitted 2026-08-03Batteryless V2X overlay boosts throughput 15%
Zhaoyu Liu +5 · “PHY-Layer Modeling and Throughput-Driven Adaptation for Batteryless V2X Networks”
2608.01928 -
cs.SE arXiv submitted 2026-08-03LLM-built dependency graphs beat static RAG by up to 60%
Zhongxin Liu +5 · “Effective and Efficient Context Retrieval via Partial Dependency Graph for Repository-Level Code Generation”
2608.01927 -
cs.AI arXiv submitted 2026-08-03Progress-ordered world model beats local predictors by 9.67 points
Zihan Liu +6 · “ProWorld: Progress-Aware Hyperbolic World Models for Long-Horizon Visual Goal Reaching”
2608.01926 -
cs.CE arXiv submitted 2026-08-03Transformer lifts neoantigen top-20 recall from 46.9% to 53.1%
Zhiyin An +5 · “TransNRank: Towards Accurate Neoantigen Ranking with Transformer”
2608.01924 -
cs.SD arXiv submitted 2026-08-03One model beats all baselines across all prompt modes and edits
Chong Jing +5 · “P-MUSE: Prompt-MIDI-Optional Model for Unified Instrumental Music Synthesis and Editing”
2608.01920 -
cs.LG arXiv submitted 2026-08-03Guided evolution lifts agent SWE-bench score 54% to 66%
Luan Zhang +11 · “HarnessCompass: Guiding Automatic Harness Evolution toward Generalizable and Effective Agent Harnesses”
2608.01918 -
cs.LG arXiv submitted 2026-08-03Risk-sensitive RL converges at the optimal 1/√n rate
Ankur Naskar +4 · “Finite-Time Analysis of Discounted Exponential-Utility Reinforcement Learning”
2608.01917 -
cs.CV arXiv submitted 2026-08-03Hybrid map plus overlap windows push neural SLAM tracking to 0.56 cm
Wenxuan Ji +5 · “CHOW-SLAM: Compact Hybrid Representation with Complementary Overlap Window Optimization for RGB-D SLAM”
2608.01914