INDEX
Every paper Pith has read and judged, in one searchable index.
-
cs.AI arXiv submitted 2026-07-31Policy-space mixing beats reward mixing for multi-goal LLMs
Ruiming Liang +8 · “Don't Mix Rewards, Mix Policies: Policy Decomposition and Optimization for Multi-Reward RL”
2607.29246 -
stat.ML arXiv submitted 2026-07-31Expected improvement is minimax-optimal
Emmanuel Vazquez +1 · “Simple-regret rates and minimax optimality of fixed-prior expected improvement in Mat\'ern and squared-exponential RKHSs”
2607.29245 -
cs.CR arXiv submitted 2026-07-31JUNO's vector votes drop ACS messages to O(n^2)
Liangrong Zhao +3 · “JUNO: Aggregated Vector Consensus for Optimal Asynchronous Common Subset”
2607.29244 -
cs.CV arXiv submitted 2026-07-31AI tool sizes aortic valves in minutes
Alessandra Zerillo +6 · “TAVI-TEC: An AI-Based Tool for Procedural Planning of Transcatheter Aortic Valve Implantation”
2607.29243 -
cs.IR arXiv submitted 2026-07-31Bandit routing steadies LLM-driven recommender tuning
Haoran Ling +7 · “RecHarness: A Bandit-Routed Agentic Harness for Self-Evolving Recommender Systems”
2607.29241 -
cs.CV arXiv submitted 2026-07-31Subtracting no-image bias lifts counterfactual accuracy 7.8 points
Kesheng Chen +2 · “When Model Priors Conflict with Visual Evidence: Mitigating Commonsense-Driven Hallucinations by Selective Prior Calibration”
2607.29240 -
cs.CL arXiv submitted 2026-07-31Tiny LoRA models match 7B on user-style rewriting
Antorweep Chakravorty · “Small Is Enough: Per-User Style Rewriting of AI-Edited Text via LoRA Adapters”
2607.29238 -
cs.CV arXiv submitted 2026-07-31Zero-parameter scene flow ranks second and holds long range
Minh-Quan Dao +3 · “CorrelationFlow: A Training-Free Geometric Approach for LiDAR Scene Flow Estimation”
2607.29237 -
math.CO arXiv submitted 2026-07-31One search counts every triangle-maximal pseudoline arrangement to n=27
Roman Parpalak +1 · “Enumeration and Classification of Triangle-Maximal Pseudoline Arrangements”
2607.29236 -
cs.RO arXiv submitted 2026-07-31Training-free feedback lifts robot success by 3 points
Peize Li +5 · “FBFM: A Training-Free Asynchronous Feedback Mechanism for Flow-Matching in World-Action Models Execution”
2607.29235 -
cs.RO arXiv submitted 2026-07-31Wearable $50 fingertip sensor lifts robot grasp success 40 points
Yongxi Liu +6 · “TacPrint: A Wearable Fingertip Tactile Sensor for Human-to-Robot Contact Reproduction”
2607.29231 -
cs.NE arXiv submitted 2026-07-31SOMA candidate steps are exact linear maps
Vojt\v{e}ch Nov\'ak +1 · “Linear Proposal Operators and Stochastic Search Geometry in SOMA and Differential Evolution”
2607.29228 -
cs.RO arXiv submitted 2026-07-31Event cameras beat RGB for teleop under bad light
Haoyu Fu +6 · “Event-Based Upper-Body Humanoid Teleoperation Under Challenging Illumination”
2607.29227 -
cs.LG arXiv submitted 2026-07-31Gaussian-process BO loses the cost-quality trade-off across 17 tasks
Panagiotis Krokidas +5 · “Frugal Bayesian Optimization: Scalable Surrogates for Data- and Resource-Limited Discovery”
2607.29225 -
cs.CV arXiv submitted 2026-07-31Beats supervised saliency models with no training on 13 benchmarks
Wenzhuo Zhao +8 · “Is It Time for the Renaissance of Salient Object Detection in the Era of MLLMs?”
2607.29222 -
cs.CR arXiv submitted 2026-07-31Masked matrix products make confidential 70B-model AI practical
James Hsin-yu Chiang +3 · “MOSAIC: Masked Outsourcing of Secure AI Computations”
2607.29221 -
cs.AI arXiv submitted 2026-07-31Hidden Minecraft rules can help agents
Jianxin Gao +8 · “MirrorCraft: Paired Evaluation under Hidden Rule Changes in Minecraft”
2607.29218 -
math.NA arXiv submitted 2026-07-31One number controls all 48 seventh-order error terms
Zhixin Huo · “Certified Seventh-Order Two-Derivative Hermite Deferred Correction via Node-Sweep Co-Design”
2607.29217 -
cs.IR arXiv submitted 2026-07-31Multimodal RL alignment lifts delivery orders by 0.55%
Jiping Liu +9 · “GALA: Generative Aligned Learning for Adaptive Multimodal Representation in the Taobao Shangou Recommender System”
2607.29213 -
cs.CL arXiv submitted 2026-07-31LLMs can learn when to quit: futile reasoning down to 1%
Xinyan Guan +7 · “Knowing When to Quit: Diagnosing and Training LLMs to Abort Futile Reasoning”
2607.29211 -
cs.LG arXiv submitted 2026-07-31Anneal the teacher signal: stable fusion beats fixed mixing everywhere
Yifan Ding +9 · “SAF-OPD: Stable Advantage Fusion for On-Policy Distillation”
2607.29209 -
cs.CV arXiv submitted 2026-07-31Dual semantic guidance lifts multi-spectral ReID from 47 to 64.6 mAP
Weixiang Zhou +6 · “Multi-Modal Object Re-Identification with Dual Semantic Guidance and Global-Local Mutual Modulation”
2607.29207 -
cs.RO arXiv submitted 2026-07-31Ellipse per drone keeps swarm safety real-time in dense forests
Alice Rosetti +4 · “MROPE: A Multi-Robot Safe Cooperative Strategy via combined Predictive Safety Filters and Ellipse-based Constraint Compression”
2607.29203 -
cs.CV arXiv submitted 2026-07-31Easy-first two-stage training sets new source-free adaptation records
Pan Liu +5 · “Domain-Division based Progressive Learning for Source-Free Domain Adaptation”
2607.29202 -
cs.CV arXiv submitted 2026-07-31One model segments 13 ultrasound organs from text prompts alone
Bo Xu +7 · “UltraSAM3: A Concept-Driven Foundation Model for Universal Ultrasound Image Segmentation”
2607.29200 -
cs.CR arXiv submitted 2026-07-31Multi-turn dialogue gives back ~20 points of guardrail safety
Haoxin An +8 · “Alignment Is Local: A Paired Diagnostic for GUI Agents under User-Side Persuasion”
2607.29199 -
cs.DC arXiv submitted 2026-07-31Checker catches every smart-home safety violation before execution
Rishabh Menezes +4 · “Knox: Fortifying Smart Spaces With Safety Guarantees”
2607.29198 -
cs.CL arXiv submitted 2026-07-31Best of 22 chatbots passes just 41% of strict long-dialogue checks
Eileen Ye +9 · “Hy-MultiTurn: A Six-Dimensional Benchmark for Deep Multi-Turn Dialogue Understanding”
2607.29196 -
eess.SY arXiv submitted 2026-07-31Software fix cuts short-circuit power rise by up to 36%
Lukas Zezula +3 · “Interturn Short Circuit Fault Mitigation in PMSMs”
2607.29195 -
cs.CV arXiv submitted 2026-07-31Local-consistency term lifts few-shot RS scene classification
Karim El Khoury +3 · “Locally Consistent Transductive Information Maximization for Few-Shot Remote Sensing Scene Classification”
2607.29192 -
cs.SC arXiv submitted 2026-07-31Exact proof settles 4x4 Dittert conjecture: uniform matrix wins
Jinhui Li +2 · “A Proof of the Dittert Conjecture in Dimension 4 via an Agent-Guided Exact Sum-of-Squares Certificate”
2607.29191 -
cs.AI arXiv submitted 2026-07-31Certify the joint tool-return neighborhood
Blaise Delattre +2 · “CAGE: Certified Authorization under Typed-Return Uncertainty for Tool-Using Agents”
2607.29190 -
cs.CL arXiv submitted 2026-07-31Automatic echo detection in migration stories peaks at r=0.38
Sakayo Toadoum Sari +7 · “Detecting Experiential Intertextuality Across Migration Routes: Beyond Surface Similarity in French Narratives”
2607.29188 -
cs.DM arXiv submitted 2026-07-31Two antistrong packing problems are NP-complete
Jiangdong Ai +3 · “The Complexity of Mixed Arc-Disjoint Spanning Subdigraphs with Antistrong Connectivity”
2608.00115 -
cs.CL arXiv submitted 2026-07-31Hidden reasoning traces sharpen AI preference judges
Sanwoo Lee +5 · “Learning Latent Reasoning Traces for Scalar Reward Models End-to-End”
2607.29185 -
eess.IV arXiv submitted 2026-07-31Deep learning predicts skull ultrasound aberrations 2,535× faster
Minju Seol +3 · “Few-shot Deep Learning for Phase-Amplitude Aberration Correction in Transcranial Focused Ultrasound”
2607.29182 -
cs.LG arXiv submitted 2026-07-31After eight passes, state machines beat frequency baselines
Andy J. Phu +4 · “SERUM: State Extraction and Refinement for User Modeling”
2607.29181 -
cs.CV arXiv submitted 2026-07-31Flow-friendly latent tops text-to-motion without VQ or AR
Yifei Zhu +5 · “MoRAE: Flow-Friendly Self-Supervised Latents for Text-to-Motion Generation”
2607.29180 -
cs.SE arXiv submitted 2026-07-31Hackathon teams verify AI output even without explicit rules
Wangyiyao Zhou +2 · “Quick Build, Careful Check? Generative AI Use in Hackathons”
2607.29178 -
cs.LG arXiv submitted 2026-07-31Behavior-aware diffusion fills utility data gaps up to 29% better
Rongchao Xu +4 · “MBDiff: Multi-view Behavior-aware Diffusion Model for Probabilistic Utility Data Imputation”
2607.29177 -
eess.SP arXiv submitted 2026-07-31Predicting latent states lifts EEG transfer accuracy 10 points
Jinhao Li +10 · “EEG-JEPA: Structured Latent Prediction for EEG Foundation Models”
2608.00114 -
cs.SE arXiv submitted 2026-07-31Execution-first training data beats query-first for tool agents
Hafsa Ouajdi +5 · “Execution-First Synthetic Tool-Use Trace Generation for LLM Agents”
2607.29175 -
cs.DB arXiv submitted 2026-07-31MERIT makes graph-ANN vector deletion as cheap as insertion
Zekai Wu +8 · “MERIT: Efficient In-Place Deletion for Dynamic Graph-Based Approximate Nearest Neighbor Indexes”
2607.29173 -
cs.DB arXiv submitted 2026-07-31Deleting vectors from ANN graphs now rivals insertion cost
Zekai Wu +8 · “MERIT: Efficient In-Place Deletion for Dynamic Graph-Based Approximate Nearest Neighbor Indexes”
2607.29173 -
cs.RO arXiv submitted 2026-07-31Planner-guided drift AI laps faster than human drivers in simulation
Sheng Zhao +3 · “Track-Guided Hierarchical Reinforcement Learning for Autonomous Vehicle Drifting with Minimum-Lap-Time Planning”
2608.00113 -
cs.RO arXiv submitted 2026-07-31Robot hits near-perfect success with no model access
Yuxin Chen +9 · “CLIFT: Turning Gemini Robotics On-Device into Humanoid Specialists via Non-Invasive Closed-Loop Iterative Fine-Tuning”
2607.29172 -
cs.RO arXiv submitted 2026-07-31Run-time guard lifts robot success from 49% to 90%
Wenda Yu +5 · “ActFovea: Runtime Safeguarding for VLA Policies via Spatiotemporal Visual-Action Consistency”
2607.29169 -
cs.CL arXiv submitted 2026-07-31N-grams beat transformers at verifying German video speakers
Oren Halvani +1 · “Authorship Verification of Transcribed German-Language Videos”
2607.29168 -
cs.CR arXiv submitted 2026-07-31Blocks memory-laundered LLM agent attacks
Jinghan Xu +4 · “Memory Provenance Laundering in LLM Agents: A Non-Amplification Firewall for Persistent Memory”
2607.29167 -
cs.CG arXiv submitted 2026-07-31Coverage beats 1-1/e for disks
Sujoy Bhore +2 · “Approximation Algorithms for Geometric Maximum Coverage”
2607.29160