archive
Every paper Pith has read. Search by title, abstract, or pith.
69392 papers in cs · page 1
-
Structural tokens let one model reason across proteins
Accurate, Interdisciplinary and Transparent Structure-property Understanding with Deep Native Structural Reasoning
-
Tiny model matches gpt-4o-mini on facts by offloading memory
Co-LMLM: Continuous-Query Limited Memory Language Models
-
Why delta-rule attention is the closest linear twin of softmax
The Key to Going Linear: Analysis-Driven Transformer Linearization
-
Acyclic subgraphs beat spanning-tree bound in FPT time
Exploiting Spanning Trees for Directed Acyclicity
-
Causal Slicing Finds Root Causes in Agent Traces
From Noisy Traces to Root Causes: Structural Trajectory Analysis and Causal Extraction for Agent Optimization
-
Fourier Decomposition Lets Robots Obey Speed Limits Without Bending Paths
SPECTRA: Context-Conditioned Spectral Movement Primitives for Robot Skill Generalization
-
Direct linear solves replace gradient steps in multi-input operator network training
Hybrid Least Squares/Gradient Descent Methods for MIONets
-
LLMs can learn to reuse reasoning steps during deployment
MILES: Modular Instruction Memory with Learnable Selection for Self-Improving LLM Reasoning
-
Simple text ensembles beat complex multimodal forecasters
Rethinking Multimodal Time-Series Forecasting Evaluation
-
Representation, not prompt design, limits SAM in medical scans
HPR-SAM: Hierarchical Probabilistic Representation Learning for Prompt-free SAM-based Medical Image Segmentation
-
Networks Can Agree on Some Dimensions and Differ on Others
Subspace Consensus of Matrix-Weighted Networks
-
LLM flight planner with memory and vision coach hits 94% validity
End-to-End LLM Flight Planning with RAG-based Memory and Multi-modal Coach Agent
-
Same AI that catches malware can now write it
Large Language Models (LLMs) and Generative AI in Cybersecurity and Privacy: A Survey of Dual-Use Risks, AI-Generated Malware, Explainability, and Defensive Strategies
-
Doubling data center load multiplies grid outage energy sevenfold
Evaluating Grid Resilience in the Era of Ever-Increasing Data Centers
-
Entropy regularizer keeps traffic simulators diverse
Flow-ERD: Agent-type Aware Flow Matching with Entropy-Regularized Distillation for Diverse Traffic Simulation
-
Step counts alone predict 21 health risks
Physical activities enable scalable foundation modelling for broad-spectrum health prediction
-
Quantum search finds reset sequences in square-root time
A quantum model for synchronizing finite state transition systems
-
ELSA3D halves 3D model cost by routing text to the right scale
ELSA3D: Elastic Semantic Anchoring for Unified 3D Understanding and Generation
-
3D-Lifted VLA Models Beat Prior Methods by 11% on Robot Tasks
Lift3D-VLA: Lifting VLA Models to 3D Geometry and Dynamics-Aware Manipulation
-
Floating Particles on Robots Render City Data in Mid-Air
Embodied Human-Robot Interaction via Acoustics: A MARL Approach with AcoustoBots for Spatial Data Physicalization
-
One model, no task-specific heads, matches vision specialists
Vision as Unified Multimodal Generation
-
4D world model beats 2D policies on bimanual robot tasks
RynnWorld-4D: 4D Embodied World Models for Robotic Manipulation
-
Diffusion model tracks 6-DoF motion from video alone
ProxyPose: 6-DoF Pose Tracking via Video-to-Video Translation
-
Rewriting captions for infrared evidence lifts CLIP retrieval by up to 12.8 points
MonoIR-RS: Infrared Remote Sensing Vision-Language Learning with CLIP and VLM Adaptation
-
Quantum moves can't beat perfect classical play in board games
Quantum combinatorial games
-
Style-transferred patches lift cross-site mammogram calcification AUC by 4-5 points
Unsupervised Domain Adaptation for Calcification Classification in Mammography Across Multi-Site Datasets
-
Oral Knowledge Could Counteract AI's Cultural Homogenization
Rethinking Indic AI from a Lens of Cultural Heritage Preservation
-
Chimp sequences so short even random parsers get 79% right
On the feasibility of dependency parsing of non-human sequences without a gold standard. Is evaluation possible in other species?
-
Splitting deep layers lets spoken AI talk and think at once
Hierarchical Acoustic-Semantic Modeling: Modality Separation and Semantic Coherence for Full-Duplex SLMs
-
Zero-Shot MLLM Framework Solves Robot Last-Mile Navigation
UniLM-Nav: A Unified Framework for Zero-Shot Last-Mile Navigation
-
Lipschitz Neural Net + Classical Observer = Provably Stable Robot Control
Neural-ESO: A Dual-Pathway Architecture for Provably Robust Learning-Based Control
-
Topology-aware attention lets 3D-LLMs reason across rooms
CAIRN: Cross-Room 3D Scene Understanding with Topology-Aware Large Multimodal Models
-
Quantum-inspired classical solvers hit a wall on high-dimensional PDEs
Quantum-inspired methods for finite-element discretizations of the high-dimensional Poisson equation
-
MILP generator keeps local blocks wired to their interfaces
GraphBU: MILP Instance Generation with Graph-Native Block Units
-
Swap any AI model, keep the same clinical routing
The Large Cancer Assistant (LCA): A Model-Agnostic Orchestration Framework for Scalable Clinical Decision Support in Oncology
-
One Satellite Ratio Maps Neighborhood Wealth Across 59 Indian Cities
Life Style Levels: Neighborhood Delineation using Geospatial Data
-
254x Citation Inflation Surfaced by Trust-Aware Graph Ranking
Trust-Aware Citation Cartel Ranking in Scholarly Knowledge Graphs
-
176M-parameter graph module matches 50-call LLM agents on KG question answering
RSF-GLLM: Bridging the Semantic Gap in Multi-Hop Knowledge Graph QA via Recurrent Soft-Flow and Decoupled LLM Generation
-
Threshold committee lets any chain's assets trade as ERC-20s
Crossroads: A Smart Contract Layer for Chain-Abstracted Assets
-
No linear-size shortcut for Vietoris-Rips on rough spaces
Lower Bounds for Approximating the Vietoris-Rips Filtration
-
8.3x KV Cache Cut Holds Quality in Long-Context LLMs
DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression
-
Grounding reasoning to physical outcomes lets 8B model beat expert
Bridging Physical Reasoning and Task Generalization via Visual Action Outcome Reasoning Alignment
-
Quantum sensors can keep secrets and Heisenberg precision
Differentially private quantum sensor networks
-
Point clouds stabilize closed-loop driving video generation
Point as Skeleton: Accumulated Point Cloud Enhanced Autoregressive Generation for Closed-Loop Autonomous Driving Simulation
-
Fighting game stripped to rock-paper-scissors core becomes RL testbed
FootsiesGym: A Fighting Game Benchmark for Two-Player Zero-Sum Imperfect-Information Games
-
Self-Powered Glasses Sensors Capture Pulse and Jaw Activity at 4 Microwatts
GlassTENG: Self-Powered Triboelectric Nanogenerator based Sensing of Pulse, Jaw, and Upper Facial Activity from Everyday Glasses
-
Learned controller beats fixed pipelines for multi-hop evidence retrieval
DynaKRAG: A Unified Framework for Learnable Evidence Control in Multi-Hop Retrieval-Augmented Generation
-
GitHub repositories mapped to industry sectors at 97% precision
Industry Classification of GitHub Repositories Using the North American Industry Classification System (NAICS)
-
Real-world multivariate data beats synthetic for time-series models
RMISC: A Large-scale Real-world Multivariate Corpus for Time Series Foundation Models
-
Hidden states reveal doomed LLM agent episodes at round one
Doomed from the Start: Early Abort of LLM Agent Episodes via a Recall-Controlled Probe Cascade