papers checked against Recognition
Pith ran a Recognition coverage check on each paper here and surfaced at least one Lean theorem in the public Recognition library that looks relevant to the paper's claims. The link is Pith's machine assessment, not a citation by the paper. Click into the paper for the full Pith review and the suggested theorem matches.
60 of 33987 papers
-
Role of (periodic as well as aperiodic) tessellations in contemporary composition. The cases of Tesselles sonores and Le Chapeau \`a douze cornes by Marisa Acu\~na
math.HO · 2026-01-21 · 3 theorem links
Aperiodic monotiles such as the Hat are translated into microtonal intervals in the compositions Tesselles sonores and Le Chapeau à douze cornes.
-
U-Mamba: Enhancing Long-range Dependency for Biomedical Image Segmentation
eess.IV · 2024-01-09 · 3 theorem links
U-Mamba is a hybrid CNN-SSM architecture that outperforms prior CNN and Transformer networks on biomedical image segmentation tasks by efficiently modeling long-range dependencies.
-
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
cs.LG · 2019-05-28 · 3 theorem links
EfficientNet scales network depth, width, and resolution uniformly via a compound coefficient to deliver state-of-the-art accuracy and efficiency on image classification.
-
XD-MAP: Cross-Modal Domain Adaptation via Semantic Parametric Maps for Scalable Training Data Generation
cs.CV · 2026-01-20 · 3 theorem links
XD-MAP generates pseudo labels for LiDAR semantic segmentation from camera images using parametric maps, improving 2D and 3D segmentation performance by up to 32.3 mIoU without manual labeling.
-
The thermal backreaction of a scalar field in de Sitter spacetime. II. Spectrum enhancement and holography
hep-th · 2026-01-22 · 3 theorem links
Thermal backreaction modifies de Sitter geometry so that late-exiting modes produce a blue-tilted spectrum n_S ~ 2 while the boundary theory matches the 3d Sp(N) model flow.
-
Diffusion Models Are Real-Time Game Engines
cs.LG · 2024-08-27 · 3 theorem links
A diffusion model trained on DOOM play sessions generates stable real-time interactive game frames at 20 FPS with quality near lossy JPEG.
-
Evolution of Mean Orbital Spacing in Planetary and Satellite Systems under Tidal Dissipation and Nebular Drag
astro-ph.EP · 2026-01-22 · 3 theorem links
Mean orbital distance ratios evolve only weakly under tidal dissipation and nebular drag, preserving information about primordial system configurations.
-
Test-Time Training Done Right
cs.LG · 2025-05-29 · 3 theorem links
Large-chunk online updates during inference let test-time training scale state capacity to 40% of model size and handle contexts up to 1M tokens without custom kernels.
-
High-Rate Quantized Matrix Multiplication I
cs.IT · 2026-01-23 · 3 theorem links
High-rate quantization theory yields accurate approximations for the distortion of absmax INT and FP schemes in generic weight-plus-activation matrix multiplication.
-
Knowledge-Guided Time-Varying Causal Inference for Arctic Sea Ice Dynamics
cs.LG · 2026-01-25 · 3 theorem links
KGCM-VAE uses physical knowledge to generate time-varying SSH treatments and MMD balancing to estimate causal effects on sea ice thickness, outperforming baselines on synthetic data.
-
GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks
cs.LG · 2025-10-05 · 3 theorem links
GDPval benchmark finds frontier AI models approaching industry experts on economically valuable tasks from high-GDP sectors, with linear performance gains over time.
-
Fibers of phase tropicalizations
math.AG · 2026-01-26 · 3 theorem links
Phase tropicalization is extended to SL_2 via new valuative tools that prove an affine Kapranov theorem and exhibit polynomial structure on graded rings of monomial valuations.
-
Cosmological analysis of the DESI DR1 Lyman alpha 1D power spectrum
astro-ph.CO · 2026-01-29 · 3 theorem links
DESI DR1 Lyman-alpha data yields Δ²★=0.379±0.032 and n★=-2.309±0.019 at k★=0.009 km⁻¹s and z=3, sharpening N_eff, α_s, and β_s constraints by factors of 1.18-1.90 when combined with other probes.
-
Nougat: Neural Optical Understanding for Academic Documents
cs.LG · 2023-08-25 · 3 theorem links
Nougat applies a visual transformer to convert academic PDFs into markup language while accurately handling mathematical content on a new scientific document dataset.
-
Diffusion Policy Policy Optimization
cs.RO · 2024-09-01 · 3 theorem links
DPPO fine-tunes diffusion policies via policy gradients and outperforms prior RL approaches for diffusion policies and PG-tuned alternatives on robot benchmarks while enabling stable training and hardware deployment.
-
Simplified State Space Layers for Sequence Modeling
cs.LG · 2022-08-09 · 3 theorem links
S5 uses a single MIMO state space model with S4-derived initialization to match S4 efficiency and reach 87.4% average accuracy on the Long Range Arena benchmark.
-
ZKBoost: Zero-Knowledge Verifiable Training for XGBoost
cs.CR · 2026-02-04 · 3 theorem links
The first zkPoT protocol for XGBoost using a generic template and VOLE instantiation, with a fixed-point version achieving accuracy within 1% of standard XGBoost.
-
NPCNet: Navigator-Driven Pseudo Text for Deep Clustering of Early Sepsis Phenotyping
cs.LG · 2026-02-03 · 3 theorem links
NPCNet is a deep clustering model that converts EHR data to pseudo texts and uses a navigator module to align sepsis phenotypes with clinical knowledge.
-
Isotonic Layer: A Unified Framework for Recommendation Calibration and Debiasing
cs.IR · 2026-02-02 · 3 theorem links
The Isotonic Layer unifies calibration and debiasing in recommendation models as a single end-to-end trainable component using learnable context embeddings for piecewise linear adjustments.
-
Theoretical Foundations of Conformal Prediction
math.ST · 2024-11-18 · 3 theorem links
The book curates and presents proofs of important existing results in conformal prediction in a unified pedagogical format with illustrations.
-
Smoothness Errors in Dynamics Models and How to Avoid Them
cs.LG · 2026-02-05 · 3 theorem links
Relaxed unitary convolutions for GNNs on meshes balance smoothness preservation with natural smoothing in dynamics, outperforming unitary convolutions and other models on PDEs and weather tasks.
-
LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics
cs.LG · 2025-11-11 · 3 theorem links
LeJEPA derives an optimal isotropic Gaussian target for embeddings and enforces it via sketched regularization to deliver scalable, heuristics-free self-supervised pretraining with 79% ImageNet linear accuracy on ViT-H/14.
-
MemoryBank: Enhancing Large Language Models with Long-Term Memory
cs.CL · 2023-05-17 · 3 theorem links
MemoryBank equips LLMs with long-term memory using Ebbinghaus-inspired updates, allowing recall and personality adaptation in chatbots like SiliconFriend.
-
MINT: Minimal Information Neuro-Symbolic Tree for Objective-Driven Knowledge-Gap Reasoning and Active Elicitation
cs.AI · 2026-02-04 · 3 theorem links
MINT combines symbolic trees with neural uncertainty estimation and LLM query curation to achieve near-expert planning performance by asking a small number of targeted questions that close knowledge gaps.
-
Deforming the Double-Scaled SYK & Reaching the Stretched Horizon From Finite Cutoff Holography
hep-th · 2026-02-05 · 3 theorem links
Deformations of the double-scaled SYK model via finite-cutoff holography produce Krylov complexity as wormhole length and realize Susskind's stretched horizon proposal through targeted T² deformations in the high-energy spectrum.
-
World Model on Million-Length Video And Language With Blockwise RingAttention
cs.LG · 2024-02-13 · 3 theorem links
Presents open-source 7B models for million-token video and language understanding via Blockwise RingAttention, setting new benchmarks in retrieval and long video tasks.
-
MiniMax-01: Scaling Foundation Models with Lightning Attention
cs.CL · 2025-01-14 · 3 theorem links
MiniMax-01 models match GPT-4o and Claude-3.5-Sonnet performance while providing 20-32 times longer context windows through lightning attention and MoE scaling.
-
Highly Adaptive Principal Component Regression
stat.ML · 2026-02-11 · 3 theorem links
PCHAL and PCHAR apply principal component reduction to the HAL basis for faster computation in high dimensions with empirical performance comparable to HAL and HAR.
-
Structural Learning Theory: A Metric-Topology Factorization Approach
cs.LG · 2026-02-08 · 3 theorem links
Structural Learning Theory defines width as the minimum number of jointly contractive low-risk cells needed to cover a learning problem, inducing a phase transition where K below width creates an irreducible error floor while K at or above width reduces to ordinary statistical learning.
-
DeepEyesV2: Toward Agentic Multimodal Model
cs.CV · 2025-11-07 · 3 theorem links
DeepEyesV2 uses a two-stage cold-start plus reinforcement learning pipeline to produce an agentic multimodal model that adaptively invokes tools and outperforms direct RL on real-world reasoning benchmarks.
-
CosyVoice 3: Towards In-the-wild Speech Generation via Scaling-up and Post-training
cs.SD · 2025-05-23 · 3 theorem links
CosyVoice 3 achieves better content consistency, speaker similarity, and prosody naturalness in zero-shot multilingual speech synthesis by scaling data to one million hours, model size to 1.5 billion parameters, and introducing a supervised multi-task speech tokenizer plus a differentiable reward模型.
-
Learning to Discover at Test Time
cs.LG · 2026-01-22 · 3 theorem links
TTT-Discover applies test-time RL to set new state-of-the-art results on math inequalities, GPU kernels, algorithm contests, and single-cell denoising using an open model and public code.
-
Learning beyond Teacher: Generalized On-Policy Distillation with Reward Extrapolation
cs.LG · 2026-02-12 · 3 theorem links
Generalized on-policy distillation with reward scaling above one (ExOPD) lets student models surpass teacher performance when merging domain experts on math and code tasks.
-
Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents
cs.AI · 2025-05-29 · 3 theorem links
The Darwin Gödel Machine evolves its own code via open-ended archive-based exploration with foundation models, raising SWE-bench performance from 20.0% to 50.0% and Polyglot from 14.2% to 30.7%.
-
Efficient Sparse Selective-Update RNNs for Long-Range Sequence Modeling
cs.LG · 2026-02-11 · 3 theorem links
suRNNs use neuron-level binary switches to update recurrent states only on informative events, matching Transformer accuracy on long-range tasks while remaining more efficient.
-
Symmetry Spans and Enforced Gaplessness
cond-mat.str-el · 2026-02-12 · 3 theorem links
Symmetry spans enforce gaplessness when a symmetry E embedded into two larger symmetries C and D has no compatible gapped phase that restricts from both.
-
Augmented Language Models: a Survey
cs.CL · 2023-02-15 · 3 theorem links
Augmented language models combine reasoning and tool-calling abilities with standard language modeling to improve performance on complex tasks while retaining core capabilities.
-
Holographic Equidistribution
hep-th · 2026-02-12 · 3 theorem links
Equidistribution of Hecke operators in large N CFT limits reduces the partition function to light-state Poincaré series with an immediate interpretation as sums over handlebody geometries.
-
FLAG Review 2024
hep-lat · 2024-11-06 · 3 theorem links
The FLAG 2024 review provides updated averages of lattice QCD determinations for quark masses, decay constants, form factors, mixing parameters, and nucleon matrix elements.
-
VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation
cs.CV · 2024-09-06 · 3 theorem links
VILA-U unifies visual understanding and generation inside one autoregressive next-token prediction model, removing separate diffusion components while claiming near state-of-the-art results.
-
Scaling Synthetic Data Creation with 1,000,000,000 Personas
cs.CL · 2024-06-28 · 3 theorem links
A curated set of one billion personas enables scalable, diverse synthetic data generation for LLM training across reasoning, instructions, knowledge, NPCs, and tools.
-
FastVGGT: Training-Free Acceleration of Visual Geometry Transformer
cs.CV · 2025-09-02 · 3 theorem links
FastVGGT achieves 4x speedup on VGGT for 1000-image inputs using training-free token merging tailored to 3D architectures while reducing error accumulation.
-
BeyondMimic: From Motion Tracking to Versatile Humanoid Control via Guided Diffusion
cs.RO · 2025-08-11 · 3 theorem links
BeyondMimic combines compact motion tracking with a unified guided latent diffusion model to master diverse agile behaviors from human demos and solve unseen downstream tasks via test-time classifier guidance.
-
VimRAG: Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph
cs.CV · 2026-02-13 · 3 theorem links
VimRAG models multimodal reasoning as a dynamic DAG and modulates visual token allocation by node topology, reaching state-of-the-art results on multimodal RAG benchmarks.
-
WebVoyager: Building an End-to-End Web Agent with Large Multimodal Models
cs.CL · 2024-01-25 · 3 theorem links
WebVoyager uses a large multimodal model to complete real-world web tasks end-to-end and reaches 59.1 percent success on a new benchmark of 15 live sites, with an automatic GPT-4V evaluator that matches human judgments 85 percent of the time.
-
Spherically symmetric black holes in Gravity from Entropy and spontaneous emission
gr-qc · 2026-02-14 · 3 theorem links
In the Gravity from Entropy framework, spherically symmetric black holes acquire r^{-4} corrections to Schwarzschild geometry, with large-mass evaporation at constant rate -β/24 and intermediate-mass loss following the classical Hawking M^{-2} scaling.
-
EchoTorrent: Towards Swift, Sustained, and Streaming Multi-Modal Video Generation
cs.CV · 2026-02-14 · 3 theorem links
EchoTorrent combines multi-teacher distillation, adaptive CFG calibration, hybrid long-tail forcing, and VAE decoder refinement to enable few-pass autoregressive streaming video generation with improved temporal consistency and audio-lip sync.
-
MemEvolve: Meta-Evolution of Agent Memory Systems
cs.CL · 2025-12-21 · 3 theorem links
MemEvolve jointly evolves agent experiential knowledge and memory architectures via a modular codebase, delivering up to 17% gains on agent benchmarks with cross-task and cross-model generalization.
-
Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation
cs.CV · 2024-10-17 · 3 theorem links
Janus decouples visual encoding into task-specific pathways inside a single autoregressive transformer to unify multimodal understanding and generation while outperforming earlier unified models.
-
villa-X: Enhancing Latent Action Modeling in Vision-Language-Action Models
cs.RO · 2025-07-31 · 3 theorem links
villa-X enhances latent action modeling in VLA models to support zero-shot action planning for unseen robot embodiments and open-vocabulary instructions, yielding better manipulation results in simulation and real-world tests.
-
The Completed SDSS-IV extended Baryon Oscillation Spectroscopic Survey: Cosmological Implications from two Decades of Spectroscopic Surveys at the Apache Point observatory
astro-ph.CO · 2020-07-17 · 3 theorem links
Final SDSS BAO and RSD data from galaxies, quasars, and Lyman-alpha forests confirm LambdaCDM, rule out dark-energy-free models at over 8 sigma, and give H0 = 68.20 +/- 0.81 km/s/Mpc in tension with local measurements.
-
Genie Envisioner: A Unified World Foundation Platform for Robotic Manipulation
cs.RO · 2025-08-07 · 3 theorem links
Genie Envisioner unifies robotic policy learning, simulation, and evaluation inside one instruction-conditioned video diffusion framework using GE-Base, GE-Act, and GE-Sim.
-
[b]=[d]-[t]+[p]: Self-supervised Speech Models Discover Phonological Vector Arithmetic
eess.AS · 2026-02-21 · 3 theorem links
Self-supervised speech models encode phonological features as linear, compositional vectors that support arithmetic operations like [b]=[d]-[t]+[p].
-
The CMA Evolution Strategy: A Tutorial
cs.LG · 2016-04-04 · 3 theorem links
The CMA-ES is a stochastic optimizer for continuous non-linear non-convex problems that adapts a multivariate normal search distribution using covariance matrix updates derived from intuitive requirements.
-
Discovering Latent Knowledge in Language Models Without Supervision
cs.CL · 2022-12-07 · 3 theorem links
An unsupervised technique extracts latent yes-no knowledge from language model activations by locating a direction that satisfies logical consistency properties, outperforming zero-shot accuracy by 4% on average across models and datasets.
-
Ovis2.5 Technical Report
cs.CV · 2025-08-15 · 3 theorem links
Ovis2.5 introduces native-resolution visual processing and reflective chain-of-thought to reach SOTA open-source multimodal performance at 9B and 2B scales on benchmarks including STEM and chart analysis.
-
Entanglement dynamics of many-body quantum states: sensitivity to system conditions and a hidden universality
quant-ph · 2026-02-22 · 3 theorem links
A single functional parameter unifies the entanglement statistics evolution for eigenstates of Hamiltonians represented by multiparametric Gaussian ensembles.
-
Quantum-metric-nematicity induced Kerr-like polarization rotation without time-reversal symmetry breaking
cond-mat.mes-hall · 2026-02-23 · 3 theorem links
Quantum metric nematicity produces incident-polarization-dependent Kerr-like polarization rotation in time-reversal symmetric nonmagnetic systems.
-
Helmlab: A Two-Space Family of Analytical, Data-Driven Color Spaces for UI Design Systems
cs.GR · 2026-02-26 · 3 theorem links
Helmlab defines MetricSpace and GenSpace as analytical data-driven color spaces that cut color difference error by 23% versus CIEDE2000 on primary benchmarks while outperforming OKLab on gradient tasks.
-
On the Entropy in Last-Mile Logistics
math.OC · 2026-02-26 · 3 theorem links
Structural entropy quantifies last-mile logistics fragmentation, showing high disorder in Amazon routes where spatial consolidation redistributes entropy between carriers and customers while temporal consolidation reduces total entropy.