papers checked against Recognition

60 of 33886 papers

• POETS: Uncertainty-Aware LLM Optimization via Compute-Efficient Policy Ensembles

  cs.LG · 2026-05-08 · 3 theorem links

  POETS uses compute-efficient LLM policy ensembles to implicitly perform KL-regularized Thompson sampling, delivering O(sqrt(T gamma_T)) regret bounds and state-of-the-art sample efficiency in scientific discovery tasks such as protein search and quantum circuit design.

• BeeVe: Unsupervised Acoustic State Discovery in Honey Bee Buzzing

  cs.SD · 2026-05-08 · 3 theorem links

  Unsupervised VQ-VAE training on PaSST embeddings discovers repeatable discrete acoustic tokens in honey bee buzzing that separate queenright from queenless conditions and identify three stable sub-states in queenless hives.

• Denoising Diffusion Probabilistic Models

  cs.LG · 2020-06-19 · 3 theorem links

  Denoising diffusion probabilistic models generate high-quality images by learning to reverse a fixed forward diffusion process, achieving FID 3.17 on CIFAR10.

• Orthogonal Polynomials, a Szeg\H{o}--Verblunsky Theorem and Baxter's Theorem on the Quaternionic Sphere

  math.CA · 2026-05-08 · 3 theorem links

  A theory of orthogonal polynomials on the quaternionic sphere is introduced via q-positive measures, extending the Szegő–Verblunsky and Baxter theorems with new matrix-valued Schur recurrences and Verblunsky formulas.

• Dissipative structures in compressible inviscid fluids

  math.AP · 2026-05-08 · 3 theorem links

  Energy accumulates and dissipates on codimension-one shock structures in compressible inviscid flows but not in incompressible ones, via fine-scale analysis of the defect measure in weak solutions.

• TAVIS: A Benchmark for Egocentric Active Vision and Anticipatory Gaze in Imitation Learning

  cs.RO · 2026-05-08 · 3 theorem links

  TAVIS is a released benchmark showing active vision improves imitation learning in a task-dependent manner, multi-task policies struggle with shifts, and imitation produces human-like anticipatory gaze.

• Warm Topological Langmuir Cyclotron Wave

  physics.plasm-ph · 2026-05-08 · 3 theorem links

  Finite temperature induces a Weyl point of charge 1 between warm Langmuir and RCP waves, predicting and numerically confirming a topological Langmuir-cyclotron edge mode absent in the cold limit.

• FAME: Forecasting Academic Impact via Continuous-Time Manifold Evolution

  cs.LG · 2026-05-08 · 3 theorem links

  FAME models scientific topic trajectories in continuous time to forecast paper impact more accurately than LLMs by aligning manuscripts with field momentum in a dynamic latent space.

• Probing Gravitational Wave Signatures from Periodic Orbits of Regular Black Holes in Asymptotically Safe Gravity

  gr-qc · 2026-05-08 · 3 theorem links

  The quantum parameter ξ in an asymptotically safe regular black hole shifts the innermost stable orbit, enhances whirl behavior in periodic geodesics, and produces amplitude-modulated millihertz gravitational-wave strains whose peak amplitude grows with ξ, placing them inside the sensitivity bands预计

• PET-Adapter: Test-Time Domain Adaptation for Full and Limited-Angle PET Image Reconstruction

  cs.CV · 2026-05-08 · 3 theorem links

  PET-Adapter adapts phantom-pretrained generative models for superior full-angle and limited-angle PET reconstruction at test time via layer-wise low-rank anatomical conditioning and OSEM warm-starting that cuts diffusion steps from 50 to 2.

• A Topological Rainbow Ramsey Theorem

  math.LO · 2026-05-05 · 3 theorem links

  It is consistent relative to large cardinals that every countable-to-one coloring c of pairs from ω₂ has a closed subset A of order type ω₁ on which c is injective.

• Weighted Distributions of Complex Multiplication Orders in Ordinary Isogeny Classes

  math.NT · 2026-05-08 · 3 theorem links

  Weighted class numbers yield explicit global distributions of endomorphism rings in ordinary isogeny classes, with natural density 1/(2h(D)) for primes admitting CM by a given order O_D via Chebotarev.

• A Time-Domain Method of Auxiliary Sources for Analyzing Transient Electromagnetic Interactions with GSTC-Modeled Metasurfacess

  physics.comp-ph · 2026-05-08 · 3 theorem links

  A time-domain MAS formulation is created by converting the frequency-domain GSTC into a causal convolution representation for transient EM analysis of metasurfaces.

• Channel-Level Relation to Attentive Aggregation with Neighborhood-Homogeneity Constraint for Point Cloud Analysis

  cs.CV · 2026-05-04 · 3 theorem links

  PointCRA reduces information loss in deep point cloud networks by treating temporal trend variation as an extra evaluation dimension alongside spatial and channel attention, guided by a neighborhood homogeneity constraint.

• Disagreement-Regularized Importance Sampling for Adversarial Label Corruption

  cs.LG · 2026-05-08 · 3 theorem links

  DR-IS selects low-contamination subsets via bounded rank-disagreement in proxy ensembles under an ε-contamination model, with O(√(log(N/δ)/K)) concentration rates that certify separation when the expectation gap Δ' is positive.

• Depth Anything 3: Recovering the Visual Space from Any Views

  cs.CV · 2025-11-13 · 3 theorem links

  DA3 recovers consistent visual geometry from arbitrary views via a vanilla DINO transformer and depth-ray target, setting new SOTA on a visual geometry benchmark while outperforming DA2 on monocular depth.

• Structural Generalization on SLOG without Hand-Written Rules

  cs.CL · 2026-04-28 · 3 theorem links

  A neural cellular automaton learns compositional rules from data alone to achieve structural generalization on the SLOG semantic parsing benchmark, reaching 67.3% accuracy and fully succeeding on 11 of 17 categories.

• Physics-informed operator learning for transferable energy-dissipative microstructure dynamics

  cond-mat.mtrl-sci · 2026-05-08 · 3 theorem links

  PFNet learns physics-informed operators to deliver accurate one-step and stable multi-step predictions of microstructure dynamics in Cahn-Hilliard coarsening and martensitic transformations across varying conditions.

• EviDep: Trustworthy Multimodal Depression Estimation via Disentangled Evidential Learning

  cs.LG · 2026-04-17 · 3 theorem links

  EviDep uses evidential learning with a Normal-Inverse-Gamma distribution, wavelet-based feature extraction, and explicit feature disentanglement to deliver accurate multimodal depression estimates with calibrated uncertainty.

• Genetic Information as a "Chord" of Chemical Oscillations: Emergence of Catalyst-RNA Systems Driven by Superposed Rhythms

  q-bio.OT · 2026-05-07 · 3 theorem links

  Simulations show that superposed Lotka-Volterra chemical oscillations can bias binary polymer elongation to produce catalytic loops and low-entropy informational sequences at higher rates than random selection.

• Convergence and Emergence of In-Context Reinforcement Learning with Chain of Thought

  cs.LG · 2026-05-08 · 3 theorem links

  With specific linear Transformer parameters, CoT generation equals iterative TD updates, yielding geometric error decay with CoT length until a context-length statistical floor, and those parameters globally minimize the pretraining loss.

• Response Time Enhances Alignment with Heterogeneous Preferences

  cs.LG · 2026-05-07 · 3 theorem links

  Response times modeled as drift-diffusion processes enable consistent estimation of population-average preferences from heterogeneous anonymous binary choices.

• The Causally Emergent Alignment Hypothesis: Causal Emergence Aligns with and Predicts Final Reward in Reinforcement Learning Agents

  cs.NE · 2026-05-07 · 3 theorem links

  Causal emergence in RL agents' internal representations predicts final reward and tracks learning progress across diverse tasks and algorithms.

• Who Prices Cognitive Labor in the Age of Agents? Compute-Anchored Wages

  cs.AI · 2026-05-07 · 3 theorem links

  AI agents convert compute capital into cognitive labor units, so on substitutable tasks the competitive human wage is bounded above by relative productivity times compute intensity times the rental rate of compute.

• Accelerated Relax-and-Round for Concave Coverage Problems

  cs.DS · 2026-05-07 · 3 theorem links

  An accelerated relax-and-round algorithm for concave coverage problems achieves Õ(mn ε^{-1}) runtime and a 0.827-approximation ratio for the logarithmic reward function.

• Rhamba: Region-Aware Hybrid Attention-Mamba Framework for Self-Supervised Learning in Resting-State fMRI

  cs.LG · 2026-05-02 · 3 theorem links

  Rhamba uses region-aware masking strategies and hybrid Attention-Mamba models pretrained on ABIDE fMRI data to achieve top AUROC on schizophrenia and ADHD classification tasks while outperforming prior methods.

• A-MEM: Agentic Memory for LLM Agents

  cs.CL · 2025-02-17 · 3 theorem links

  A-MEM is a dynamic memory system for LLM agents that builds and refines an interconnected network of notes with agent-driven linking and evolution, showing performance gains over prior memory methods on six models.

• Steering Language Models With Activation Engineering

  cs.CL · 2023-08-20 · 3 theorem links

  Activation Addition steers language models by adding contrastive activation vectors from prompt pairs to control high-level properties like sentiment and toxicity at inference time without training.

• Anti De Sitter Space And Holography

  hep-th · 1998-02-20 · 3 theorem links

  Large-N limits of certain conformal field theories are equivalent to supergravity on AdS space, with CFT correlation functions obtained from the supergravity action evaluated on boundary asymptotics and operator dimensions equal to particle masses.

• LLaVA-Video: Video Instruction Tuning With Synthetic Data

  cs.CV · 2024-10-03 · 3 theorem links

  LLaVA-Video-178K is a new synthetic video instruction dataset that, when combined with existing data to train LLaVA-Video, produces strong results on video understanding benchmarks.

• Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory

  cs.CL · 2025-04-28 · 3 theorem links

  Mem0 improves long-term LLM conversational performance by up to 26% on LLM-as-Judge while cutting p95 latency 91% and token costs over 90% versus full-context baselines.

• Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets

  cs.CV · 2023-11-25 · 3 theorem links

  Stable Video Diffusion scales latent video diffusion models via text-to-image pretraining, video pretraining on curated data, and high-quality finetuning to produce competitive text-to-video and image-to-video results while enabling motion LoRA and multi-view 3D applications.

• Fine-Tuning Language Models from Human Preferences

  cs.CL · 2019-09-18 · 3 theorem links

  Language models fine-tuned via RL on 5k-60k human preference comparisons produce stylistically better text continuations and human-preferred summaries that sometimes copy input sentences.

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks

  cs.CL · 2020-05-22 · 3 theorem links

  RAG models set new state-of-the-art results on open-domain QA by retrieving Wikipedia passages and conditioning a generative model on them, while also producing more factual text than parametric baselines.

• Soft Tournament Equilibrium

  cs.AI · 2026-04-06 · 3 theorem links

  STE is a differentiable method to compute continuous analogues of the Top Cycle and Uncovered Set from pairwise comparison data for stable set-valued evaluation of cyclic agent interactions.

• SGANet: Semantic and Geometric Alignment for Multimodal Multi-view Anomaly Detection

  cs.CV · 2026-04-07 · 3 theorem links

  SGANet uses selective cross-view feature refinement, semantic-structural patch alignment, and multi-view geometric alignment to achieve state-of-the-art anomaly detection and localization on multimodal multi-view datasets.

• Settled Elements in Arboreal Galois Groups of Quadratic PCF Polynomials

  math.NT · 2026-04-06 · 3 theorem links

  Arithmetic iterated monodromy groups of postcritically finite quadratic polynomials with periodic postcritical orbits are densely settled.

• SignalClaw: LLM-Guided Evolutionary Synthesis of Interpretable Traffic Signal Control Skills

  cs.AI · 2026-04-07 · 3 theorem links

  SignalClaw synthesizes interpretable, composable traffic signal control skills through LLM-guided evolution that matches top baselines on routine SUMO scenarios and outperforms them on emergency and transit events while remaining editable by engineers.

• The Topology of Multimodal Fusion: Why Current Architectures Fail at Creative Cognition

  cs.AI · 2026-04-06 · 3 theorem links

  Multimodal AI architectures share a geometric prior of modal separability termed contact topology that prevents creative cognition, addressable via a cruciform philosophical-mathematical framework using fiber bundles and Neural ODEs.

• A Demon that remembers: An agential approach towards quantum thermodynamics of temporal correlations

  quant-ph · 2026-04-06 · 3 theorem links

  A classical agent extracts more work from quantum temporal correlations via adaptive strategies bounded by the new Time-Ordered Free Energy, while reinforcement learning achieves polylogarithmic dissipation when learning unknown states.

• Training Transformers in Cosine Coefficient Space

  cs.PF · 2026-04-06 · 3 theorem links

  Training transformers by optimizing only half the DCT coefficients per linear layer achieves validation loss within 0.024 of a dense baseline on Shakespeare character prediction, outperforming matched-parameter LoRA due to preserved rank flexibility.

• Domain-Contextualized Inference: A Computable Graph Architecture for Explicit-Domain Reasoning

  cs.AI · 2026-04-06 · 3 theorem links

  A five-layer computable graph architecture makes domain an explicit first-class parameter, yielding domain-scoped pruning, substrate-agnostic operations, and transparent inference with reliability conditions C1-C4.

• Graph-Based Chain-of-Thought Pruning for Reducing Redundant Reflections in Reasoning LLMs

  cs.CL · 2026-04-07 · 3 theorem links

  Graph-based pruning of redundant reflections in LLM chain-of-thought reduces average reasoning tokens by 42% while preserving or improving accuracy.

• HybridKV: Hybrid KV Cache Compression for Efficient Multimodal Large Language Model Inference

  cs.AI · 2026-04-07 · 3 theorem links

  HybridKV reduces KV cache memory by up to 7.9x and speeds decoding by 1.52x in MLLMs with almost no performance loss by classifying heads into static and dynamic types and compressing them differently.

• Causal Dynamical Triangulations: New Lattice Theory of Quantum Gravity

  hep-th · 2026-04-07 · 3 theorem links

  Causal Dynamical Triangulations regularizes the gravitational path integral via causal triangulations, enabling Monte Carlo simulations that produce an emergent de Sitter universe with short-distance spectral dimension near 2 and evidence for an ultraviolet fixed point.

• Incremental Risk Assessment for Cascading Failures in Large-Scale Multi-Agent Systems

  eess.SY · 2026-04-07 · 3 theorem links

  A risk-assessment framework for cascading failures in time-delay consensus networks provides closed-form expressions depending on network spectrum and delays, plus lower bounds on achievable performance.

• The Bott Metric: A Real-Space Bridge Between Topology and Quantum Metric

  cond-mat.dis-nn · 2026-04-06 · 3 theorem links

  The Bott metric, derived from the plaquette operator, unifies topology and quantum geometry by converging to the trace of the integrated quantum metric in the thermodynamic limit for disordered and amorphous systems.

• Free-Range Gaussians: Non-Grid-Aligned Generative 3D Gaussian Reconstruction

  cs.CV · 2026-04-06 · 3 theorem links

  Free-Range Gaussians uses flow matching over Gaussian parameters to predict non-grid-aligned 3D Gaussians from multi-view images, enabling synthesis of plausible content in unobserved regions with fewer primitives than grid-aligned methods.

• Decocted Experience Improves Test-Time Inference in LLM Agents

  cs.AI · 2026-04-06 · 3 theorem links

  Decocted experience—extracting and organizing the essence from accumulated interactions—enables more effective context construction that improves test-time inference in LLM agents on math, web, and software tasks.

• SuperLocalMemory V3.3: The Living Brain -- Biologically-Inspired Forgetting, Cognitive Quantization, and Multi-Channel Retrieval for Zero-LLM Agent Memory Systems

  cs.AI · 2026-04-06 · 3 theorem links

  SuperLocalMemory V3.3 implements a cognitive memory taxonomy with mathematical forgetting and multi-channel retrieval, reaching 70.4% on LoCoMo in zero-LLM mode.

• Intercity mobility reveals the hyperbolic geometry of city systems

  physics.soc-ph · 2026-04-06 · 3 theorem links

  Intercity mobility embeds into hyperbolic geometry, revealing bottom-up city hierarchies and trade-offs between hierarchy and hinterland range that shape urban structures.

• Phase-Associative Memory: Sequence Modeling in Complex Hilbert Space

  cs.CL · 2026-04-06 · 3 theorem links

  PAM, a complex-valued associative memory model, exhibits steeper power-law scaling in loss and perplexity than a matched real-valued baseline when trained on WikiText-103 from 5M to 100M parameters.

• Families of periodic solutions of the 4- and 6-body problem using a gradient-free continuation method

  math.DS · 2026-04-07 · 3 theorem links

  A gradient-free stochastic continuation method constructs families of pseudo-periodic planar solutions for symmetric 4- and 6-body problems.

• Task Ecologies and the Evolution of World-Tracking Representations in Large Language Models

  stat.ME · 2026-04-07 · 3 theorem links

  Autoregressive next-token learning selects for world-tracking representations in LLMs if and only if the encoding preserves the training ecology's equivalence classes, as shown by decomposing cross-entropy into irreducible conditional entropy plus a Jensen-Shannon excess that vanishes exactly on the

• Gradual Cognitive Externalization: From Modeling Cognition to Constituting It

  cs.AI · 2026-04-06 · 3 theorem links

  Ambient AI systems transition from modeling cognition to constituting part of users' cognitive architectures through sustained causal coupling, under a functionalist view and the no behaviorally invisible residual hypothesis.

• PASTA: Vision Transformer Patch Aggregation for Weakly Supervised Target and Anomaly Segmentation

  cs.CV · 2026-04-07 · 3 theorem links

  PASTA uses ViT feature distribution analysis and SAM to achieve up to 88.3% target and 63.5% anomaly IoU with 75.8% reduced training time under weak image-level supervision on custom datasets.

• A Frame is Worth One Token: Efficient Generative World Modeling with Delta Tokens

  cs.CV · 2026-04-06 · 3 theorem links

  Delta tokens compress VFM feature differences into single tokens, enabling a lightweight generative world model that predicts diverse futures with far lower compute than existing approaches.

• 3D-Fixer: Coarse-to-Fine In-place Completion for 3D Scenes from a Single Image

  cs.CV · 2026-04-06 · 3 theorem links

  3D-Fixer performs in-place 3D asset completion from single-view partial point clouds via coarse-to-fine generation with ORFA conditioning, plus a new ARSG-110K dataset, to achieve higher geometric accuracy than MIDI and Gen3DSR while keeping diffusion efficiency.

• Novel Light-Induced States in Triangular Metallic Magnet

  cond-mat.str-el · 2026-04-07 · 3 theorem links

  Circularly polarized light induces vortex states and long-range magnetic orders at Gamma, K/2, K, and M points in a triangular-lattice double-exchange model.

• Quantum Algorithms for Heterogeneous PDEs: The Neutron Diffusion Eigenvalue Problem

  quant-ph · 2026-04-06 · 3 theorem links

  A hybrid quantum algorithm solves the neutron diffusion eigenvalue problem for heterogeneous media with claimed polynomial end-to-end speedup over classical methods.

next page →

← previous page