archive
Every paper Pith has read. Search by title, abstract, or pith.
125528 papers indexed · page 1
-
LLM-planned 3D paths keep object identities stable in long videos
WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory
-
Three alignments convert any input into consistent 4D
Alignment Is All You Need For X-to-4D Generation
-
Plain pixel diffusion on point maps beats latent models for 3D geometry
PointDiT: Pixel-Space Diffusion for Monocular Geometry Estimation
-
Stateful monitor halves gradual-attack evasion in persistent AI codebases
Distributed Attacks in Persistent-State AI Control
-
Testbed shows unlearning often misses the parameters holding data
LACUNA: A Testbed for Evaluating Localization Precision for LLM Unlearning
-
Compile fuzzy functions into 23 MB weights
Program-as-Weights: A Programming Paradigm for Fuzzy Functions
-
Normalizing series to probabilities turns quotient signs into moment checks
A Probabilistic Sign Rule for Quotients of Positive Series and Integral Transforms
-
Simple threshold monitor matches advanced LLM safety checks
Online Safety Monitoring for LLMs
-
Evidence replay lifts long-context reasoning in LLMs without training
ReContext: Recursive Evidence Replay as LLM Harness for Long-Context Reasoning
-
Blocking cross-noise attention leaves or raises diffusion model performance
From SRA to Self-Flow: Data Augmentation or Self-Supervision?
-
Social context produces 40% public-private split in LLM agents
What LLM Agents Say When No One Is Watching: Social Structure and Latent Objective Emergence in Multi-Agent Debates
-
Integrability-breaking gates trigger chaos via OTOC hotspots
On the emergence of quantum many-body chaos for tunably-broken integrability
-
PTA data disfavors phase transitions near EFT boundary
A critical look at low-scale cosmological phase transitions in the PTA era
-
Reasoning model boosts speaker ID accuracy in TV dramas
Reasoning LLM Improves Speaker Recognition in Long-form TV Dramas
-
Tactile deformation prediction lifts contact task success to 71.67%
VT-WAM: Visual-Tactile World Action Model for Contact-Rich Manipulation
-
DemoPSD reduces leakage in LLM self-distillation via adaptive barycenters
DemoPSD: Disagreement-Modulated Policy Self-Distillation
-
C++ runtime runs embodied models on any robot hardware
Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots
-
Aharoni-Korman conjecture holds for countable FAC posets without forbidden chains
The structure of FAC posets and the Aharoni--Korman conjecture
-
SOAP and SOAP-Muon beat Adam on ML interatomic potential training
Beyond Adam: SOAP and Muon for Faster, Label-Efficient Training of Machine Learning Interatomic Potentials
-
Jeffreys prior recenters DESI DR1 posteriors for H0
Alleviating prior dependencies for DESI DR1 clustering fits through reparameterization
-
Spatial memory guides efficient 360° object search and segmentation
Seek to Segment: Active Perception for Panoramic Referring Segmentation
-
Behavior latents add independent speed and safety control to traffic sims
Controllable Sim Agents with Behavior Latents
-
AdS field Lagrangians reduce to algebra
BRST-BV approach to fields in Poincare patch of AdS
-
Tuning a few attention heads blocks text attacks on vision models
Towards Robustness against Typographic Attack with Training-free Concept Localization
-
Self-similar wave solutions stay stable forward in time
Stability of global self-similar solutions to the cubic wave equation and the wave maps equation
-
Jastrow factor raises wave function regularity order from 1 to 2
Cut-off Jastrow Factors and Spectral Barron Regularity of Coulombic Electronic Wave Functions
-
Neural guidance zeros conflicts in Sudoku solvers when hints can be overwritten
G-RRM: Guiding Symbolic Solvers with Recurrent Reasoning Models
-
Masked prefixes and replay boost VLM accuracy on new images
Visually Grounded Self-Reflection for Vision-Language Models via Reinforcement Learning
-
Almost supermartingales obey Olivier's convergence rate
Almost Supermartingale Extensions of Olivier's Theorem
-
Ipseome dataset releases largest free open human identity data
Building the Ipseome: Large, Free, Open, Human Identity Data
-
β PsA luminosity overstated 48% as pole-on rapid rotator
The Debris Disk Host $\beta$ Piscis Austrinus is a Rapidly Rotating Star Seen Nearly Pole-On
-
Descriptor-free localization cuts rotation error by 89%
GeoMix: Descriptor-Free Visual Localization via Global Context and Multi-Detector Training
-
Pi Σ-bar interaction removes need for Σ(1380) in J/ψ decay
Identifying $\Sigma(1380)$ and $\Sigma(1430)$ in the $J/\psi \to \Lambda \pi \bar{\Sigma}$ reaction
-
Entropy filtering removes noise so VLMs keep detail with fewer tokens
Combating Textual Noise and Redundancy: Entropy-Aware Dense Visual Token Pruning
-
Random percolation preserves cycles of length nearly d
Robustness and hyperstability for the Erd\H{o}s-Gallai theorem
-
Chain maps automate logical CNOTs between different CSS codes
Automated logical Clifford gadgets for heterogeneous architectures via chain maps
-
Noise symmetries fix gauge using error type alone
Symmetries of Pauli Noise from Lindbladian Dynamics
-
Merger stretches Andromeda disc beyond 40 kpc
The Merger-Driven Origin of the Vast Extended Stellar Disc Around the Andromeda Galaxy
-
Global priors raise 360° search accuracy nearly eightfold
EAGLE-360: Embodied Active Global-to-Local Exploration in 360$^\circ$
-
Sliding mode law docks 3D vehicles using range and sight angles
Docking of Autonomous Vehicles with a Stationary Docking Station in 3D Space
-
Subcritical percolation gives spin mixing time log N / lambda
Mixing times of spin systems on dynamical percolation
-
Prototype niobium cavity reaches 11 kHz mode split at 2 K
Cryogenic RF characterization of the MAGO cavity for high-frequency gravitational-wave detection
-
Geodesic equation on Hermitian manifolds gains C^{1,1} solutions
Regularity of a Geodesic equation in the space of mixed Volume Forms on Hermitian Manifolds
-
QuadRocket achieves almost global trajectory tracking with adaptive control
QuadRocket: An Aerial Robotic Testbed for Adaptive Thrust-Vector Control of Rocket-Like Vehicles
-
Narration acoustics predict audiobook appeal beyond title
Audio-Based Understanding of Audiobook Narration Appeal
-
Quadrotor learns to intercept using only camera direction vectors
Learning Agile Intruder Interception using Differentiable Quadrotor Dynamics
-
Anchored flow removes clouds while keeping semantics intact
Interpretation-Oriented Cloud Removal via Observation-Anchored Residual Flow with Geo-Contextual Alignment
-
Microwave shielding on n=1 to 2 avoids bound states
Microwave shielding of ultracold polar molecules on the transition $\boldsymbol{n=1 \rightarrow 2}$
-
Benchmark tracks test changes after code commits
TestEvo-Bench: An Executable and Live Benchmark for Test and Code Co-Evolution
-
Clifford torus is Willmore for every Berger sphere parameter
Bifurcations of the Clifford Torus as Willmore Surfaces in Berger Spheres