papers checked against Recognition

60 of 33987 papers

• Linear Geometry: flats, ranks, regularity, parallelity

  math.HO · 2025-11-20 · 3 theorem links

  A survey of foundational concepts in Linear Geometry including flats, ranks, regularity, modularity, and parallelity all derived from flat hulls.

• On the Bergman Kernel of complex hyperbolic manifolds

  math.DG · 2025-11-20 · 3 theorem links

  The Bergman kernel on polarized complex hyperbolic manifolds equals a sum over geodesic loops, yielding max/min results and off-diagonal estimates.

• Quantum simulation of out-of-equilibrium dynamics in gauge theories

  quant-ph · 2025-09-03 · 3 theorem links

  The paper reviews advances in quantum simulation of out-of-equilibrium dynamics in gauge theories, covering particle production, string breaking, thermalization, and related phenomena.

• H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models

  cs.LG · 2023-06-24 · 3 theorem links

  H2O evicts non-heavy-hitter tokens from the KV cache using a dynamic submodular policy, retaining recent and frequent-co-occurrence tokens to reduce memory while preserving accuracy.

• Temporal Graph Networks for Deep Learning on Dynamic Graphs

  cs.LG · 2020-06-18 · 3 theorem links

  Temporal Graph Networks combine memory modules and graph operators to learn on dynamic graphs as timed event sequences, outperforming prior methods on transductive and inductive tasks while unifying earlier models as special cases.

• Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs

  cs.AI · 2024-10-24 · 3 theorem links

  Data-centric filtering yields an 80K preference dataset and reward models that lead RewardBench while boosting other top entries.

• WebSailor: Navigating Super-human Reasoning for Web Agent

  cs.CL · 2025-07-03 · 3 theorem links

  WebSailor trains open-source web agents to match proprietary performance on complex information-seeking tasks by generating high-uncertainty scenarios and using a new RL method called DUPO.

• Recursively Summarizing Books with Human Feedback

  cs.CL · 2021-09-22 · 3 theorem links

  Recursive decomposition plus human feedback lets GPT-3 produce book-length summaries that reach human quality on a few cases and set new records on BookSum and NarrativeQA.

• MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory

  cs.CL · 2026-01-06 · 3 theorem links

  MemRL enables self-evolving AI agents through reinforcement learning on episodic memory with a two-phase retrieval process that filters noise and selects high-utility strategies based on environmental feedback.

• Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models

  cs.AI · 2024-06-14 · 3 theorem links

  LLMs trained on simple specification gaming generalize to zero-shot reward tampering including rewriting their own reward function.

• Lectures on Generalized Symmetries

  hep-th · 2023-07-14 · 3 theorem links

  Lecture notes that systematically introduce higher-form symmetries, SymTFTs, higher-group symmetries, and related concepts in QFT using gauge theory examples.

• Training Language Models to Self-Correct via Reinforcement Learning

  cs.LG · 2024-09-19 · 3 theorem links

  SCoRe uses multi-turn online RL with regularization on self-generated traces to improve LLM self-correction, achieving 15.6% and 9.1% gains on MATH and HumanEval for Gemini models.

• Black hole chemistry: thermodynamics with Lambda

  hep-th · 2016-08-22 · 3 theorem links

  Treating the cosmological constant as pressure in black hole thermodynamics yields an extended dictionary with enthalpy, thermodynamic volume, and chemical-like phase transitions including Van der Waals behavior, reentrant transitions, and triple points.

• 3D Reconstruction with Spatial Memory

  cs.CV · 2024-08-28 · 3 theorem links

  Spann3R uses a learned spatial memory to regress per-image pointmaps directly in a shared global coordinate system, removing the need for optimization-based alignment after per-pair predictions.

• Mogao: An Omni Foundation Model for Interleaved Multi-Modal Generation

  cs.CV · 2025-05-08 · 3 theorem links

  Mogao presents a causal unified model with deep fusion, dual encoders, and interleaved position embeddings that achieves strong performance on multi-modal understanding, text-to-image generation, and coherent interleaved outputs including zero-shot editing.

• Inflation in theories with broken diffeomorphisms

  gr-qc · 2025-11-24 · 3 theorem links

  Breaking diffeomorphism invariance to transverse diffeomorphisms in the inflaton sector yields modified slow-roll parameters and power spectra with novel post-inflationary regimes for quadratic potentials.

• The local Turnpike Property in Mean Field Control and Games with quadratic Hamiltonian

  math.AP · 2025-11-24 · 3 theorem links

  Under a local second-order strict positivity condition on stationary equilibria, mean field game systems with quadratic Hamiltonians exhibit an exponential turnpike property near those equilibria on tori and Euclidean space, with existence of stable solutions established via fixed-point methods.

• Hopfield Networks is All You Need

  cs.NE · 2020-07-16 · 3 theorem links

  Modern Hopfield networks store exponentially many patterns, retrieve them in one update, and have an update rule equivalent to transformer attention, enabling new Hopfield layers that improve results on multiple instance learning and drug design tasks.

• Video-Holmes: Can MLLM Think Like Holmes for Complex Video Reasoning?

  cs.CV · 2025-05-27 · 3 theorem links

  Video-Holmes benchmark shows top MLLMs achieve at most 45% accuracy on tasks needing integration of multiple clues from suspense films, unlike existing perception-focused tests.

• Parallel matching-based AMG preconditioners for elliptic equations discretized by IgA

  math.NA · 2025-11-26 · 3 theorem links

  Matching-based AMG preconditioners deliver robust and scalable performance for solving large ill-conditioned systems from IgA discretizations in parallel HPC settings.

• Agent Laboratory: Using LLM Agents as Research Assistants

  cs.HC · 2025-01-08 · 3 theorem links

  Agent Laboratory is an autonomous LLM framework that completes end-to-end research from idea to report and code, with human feedback improving quality and cutting expenses by 84% while reaching competitive ML performance.

• Language-driven Semantic Segmentation

  cs.CV · 2022-01-10 · 3 theorem links

  LSeg achieves competitive zero-shot semantic segmentation by contrastively aligning dense pixel embeddings from a transformer with text embeddings of class labels.

• Time-R1: Post-Training Large Vision Language Model for Temporal Video Grounding

  cs.CV · 2025-03-17 · 3 theorem links

  Time-R1 applies RL with verifiable rewards to post-train LVLMs for temporal video grounding, reaching state-of-the-art results on multiple datasets using only 2.5K samples while also improving general video capabilities.

• Primary gravitational waves at high frequencies I: Origin of suppression in the power spectrum

  astro-ph.CO · 2025-12-03 · 3 theorem links

  Adiabatic regularization combined with smoothed transitions suppresses the high-frequency oscillations in the power spectrum of primary gravitational waves about a zero mean.

• Mercury: Ultra-Fast Language Models Based on Diffusion

  cs.CL · 2025-06-17 · 3 theorem links

  Mercury Coder diffusion LLMs achieve throughputs of 1109 and 737 tokens per second on H100 GPUs, up to 10x faster than frontier models with comparable quality.

• Chukchi Myths perspective on Special Relativity

  physics.gen-ph · 2025-12-03 · 3 theorem links

  Chukchi mythology provides a perspective for teaching special relativity that prioritizes absolute structures such as proper time over relative ones to reduce the sense of revolutionary mystery.

• From RAG to Memory: Non-Parametric Continual Learning for Large Language Models

  cs.CL · 2025-02-20 · 3 theorem links

  HippoRAG 2 improves on standard RAG and prior HippoRAG by adding deeper passage integration and more effective LLM use in Personalized PageRank, delivering superior performance on factual, sense-making, and associative memory tasks including a 7% gain in associative memory over state-of-the-art.

• A new initialisation to Control Gradients in Sinusoidal Neural network

  cs.LG · 2025-12-06 · 3 theorem links

  A closed-form initialization for SIREN networks based on pre-activation fixed points and Jacobian variance sequences improves gradient scaling, training dynamics via NTK, and generalization on reconstruction tasks over the original scheme.

• Reinterpreting Landauer conductance, solving the quantum measurement problem, grand unification

  cond-mat.mes-hall · 2025-12-10 · 3 theorem links

  Negative local partial density of states reinterprets Landauer conductance as deterministic quantum measurement, solving the measurement problem and unifying classical-quantum laws.

• Gravitational-wave parameter estimation to the Moon and back: massive binaries and the case of GW231123

  gr-qc · 2025-12-10 · 3 theorem links

  LGWA could observe more than one third of known binary black hole events, detect ~90 mergers per year, and measure chirp mass better than third-generation detectors for massive systems.

• Ragas: Automated Evaluation of Retrieval Augmented Generation

  cs.CL · 2023-09-26 · 3 theorem links

  Ragas supplies reference-free metrics for measuring context relevance, faithfulness to retrieved passages, and answer quality in RAG pipelines.

• High-Resolution Measurements with the CTAO Southern Array: The Case for Pulsar Wind Nebulae

  astro-ph.HE · 2025-12-19 · 3 theorem links

  Simulations show CTAO southern array observations at multi-TeV energies, combined with X-ray data, can likely constrain magnetic field and electron distributions in PWNe such as HESS J1813-178 and MSH 15-52, with gains from better resolution offset by photon statistics limits.

• Planck 2018 results. I. Overview and the cosmological legacy of Planck

  astro-ph.CO · 2018-07-17 · 3 theorem links

  Planck 2018 provides the tightest constraints to date on the six parameters of the Lambda-CDM cosmological model from cosmic microwave background temperature and polarization data.

• Superradiance -- the 2020 Edition

  gr-qc · 2015-01-26 · 3 theorem links

  Black-hole superradiance extracts energy via the ergoregion and can trigger instabilities with applications to dark matter, beyond-Standard-Model physics, and laboratory analogs.

• Declarative distributed algorithms as axiomatic theories in three-valued modal logic over semitopologies

  cs.LO · 2025-12-24 · 3 theorem links

  Distributed algorithms are expressed as axiomatic theories in three-valued modal logic over semitopologies, with examples including Bracha broadcast and Crusader agreement, and all proofs formalized in Lean 4.

• Finite Groups of Random Walks in the Quarter Plane and Periodic $4$-bar Links

  math.AG · 2025-12-26 · 3 theorem links

  Explicit necessary and sufficient conditions are derived for finite groups of order 2n in quarter-plane random walks and for n-periodic Darboux transformations of 4-bar links, for every n greater than or equal to 2.

• Hallucination Detection and Evaluation of Large Language Model

  cs.CL · 2025-12-27 · 3 theorem links

  HHEM delivers fast hallucination detection in LLMs via classification, cutting evaluation time from 8 hours to 10 minutes with up to 82.2% accuracy while adding segment retrieval for summarization.

• Cognitive Architectures for Language Agents

  cs.AI · 2023-09-05 · 3 theorem links

  CoALA is a modular cognitive architecture for language agents that organizes memory components, action spaces for internal and external interaction, and a generalized decision-making loop to support more systematic development of capable agents.

• Mixture-of-Agents Enhances Large Language Model Capabilities

  cs.CL · 2024-06-07 · 3 theorem links

  A layered Mixture-of-Agents system combining multiple LLMs achieves state-of-the-art results on AlpacaEval 2.0 (65.1%), MT-Bench, and FLASK, outperforming GPT-4 Omni.

• PRAXIS: Integrating Program Analysis with Observability for Root-Cause Analysis

  cs.DC · 2025-12-26 · 3 theorem links

  PRAXIS combines LLM-driven structured traversal of service dependency graphs and hammock-block program dependence graphs to improve root-cause analysis accuracy by up to 6.3x while cutting token consumption by 5.3x on 30 real-world cloud incidents.

• On Geometric Evolution and Microlocal Regularity of the Navier-Stokes Equations

  physics.gen-ph · 2026-01-02 · 3 theorem links

  A microlocal lift of Navier-Stokes dynamics on manifolds yields an if-and-only-if geometric criterion for solution blow-up in terms of deformation integrability, directional entropy, and lifted energy.

• Reward Forcing: Efficient Streaming Video Generation with Rewarded Distribution Matching Distillation

  cs.CV · 2025-12-04 · 3 theorem links

  Reward Forcing combines EMA-Sink tokens and Rewarded Distribution Matching Distillation to deliver state-of-the-art streaming video generation at 23.1 FPS without copying initial frames.

• A decoder-only foundation model for time-series forecasting

  cs.CL · 2023-10-14 · 3 theorem links

  A pretrained decoder-only patched transformer achieves near state-of-the-art zero-shot forecasting performance across diverse time series datasets and settings.

• Mem-{\alpha}: Learning Memory Construction via Reinforcement Learning

  cs.CL · 2025-09-30 · 3 theorem links

  Mem-alpha uses RL to train agents on memory construction via QA-based rewards, yielding better performance than baselines and generalization from 30k to over 400k tokens.

• TellWhisper: Tell Whisper Who Speaks When

  eess.AS · 2026-01-07 · 3 theorem links

  TellWhisper jointly models speaker identity and temporal structure inside the speech encoder via TS-RoPE and Hyper-SD to improve multi-speaker ASR under overlap and fast turn-taking.

• Hybrid Quantum-Classical Ridgelet Neural Networks for Portfolio Optimization

  cs.LG · 2026-01-07 · 3 theorem links

  A hybrid model uses ridgelet feature extraction inside quantum neural networks for stock time-series forecasting and QAOA to select optimal portfolios from the forecasts.

• The Art of Scaling Reinforcement Learning Compute for LLMs

  cs.LG · 2025-10-15 · 3 theorem links

  A 400k+ GPU-hour study shows RL scaling in LLMs follows predictable sigmoidal trajectories, with most design choices affecting efficiency rather than the performance asymptote, enabling accurate large-scale predictions via the ScaleRL recipe.

• Aligned explanations in neural networks

  cs.LG · 2026-01-07 · 3 theorem links

  PiNets are pseudo-linear networks that form linear models instance-wise to enforce explanatory alignment, where explanations construct rather than rationalize predictions, and are evaluated as faithful on image classification and segmentation via MARS criteria.

• Quantum mechanics provides the physical basis of teleological evolutions

  physics.gen-ph · 2026-01-08 · 3 theorem links

  Quantum mechanics supplies the physical basis for teleological evolutions in both quantum algorithms and living systems through future attractors.

• Non-linear parabolic PDEs with rough coefficients and critical data: existence, uniqueness and regularity of weak solutions

  math.AP · 2026-01-08 · 3 theorem links

  Existence and uniqueness of maximal weak solutions to non-linear parabolic PDEs with rough coefficients and critical Besov data are proved in weighted Z-spaces without smallness conditions via novel hypercontractive SIO theory.

• LAB-Bench: Measuring Capabilities of Language Models for Biology Research

  cs.AI · 2024-07-14 · 3 theorem links

  LAB-Bench provides over 2,400 multiple-choice questions to measure LLM performance on real biology research tasks like literature recall, figure reading, database access, and sequence manipulation, with initial results compared against human expert biologists.

• On Causality and Predictivity

  gr-qc · 2026-01-09 · 3 theorem links

  Causality is a misleading metaphysical concept that should be abandoned in fundamental physics, making renouncing microcausality acceptable for consistent quantum gravity theories such as those using fakeons.

• R1-VL: Learning to Reason with Multimodal Large Language Models via Step-wise Group Relative Policy Optimization

  cs.AI · 2025-03-17 · 3 theorem links

  R1-VL uses StepGRPO with rule-based StepRAR and StepRVR rewards to let MLLMs learn step-by-step reasoning beyond imitation of positive paths.

• Krylov Complexity

  hep-th · 2025-07-08 · 3 theorem links

  Krylov complexity is a canonical, parameter-independent measure of operator spreading that probes chaotic dynamics to late times and admits a geometric interpretation in holographic duals.

• OLMoE: Open Mixture-of-Experts Language Models

  cs.CL · 2024-09-03 · 3 theorem links

  OLMoE-1B-7B is an open MoE language model activating 1B parameters per token that outperforms models with similar active parameters after pretraining on 5T tokens.

• A Roadmap to Pluralistic Alignment

  cs.AI · 2024-02-07 · 3 theorem links

  The paper formalizes three types of pluralistic AI models and three benchmark classes, arguing that current alignment techniques may reduce rather than increase distributional pluralism.

• Computer Generation of Disordered Networks with Targeted Structural Properties

  cond-mat.dis-nn · 2026-01-15 · 3 theorem links

  A bond-repulsion extension to the Wooten-Weaire-Winer algorithm plus a feedforward neural network allows efficient generation of disordered networks with targeted coordination and disorder metrics, demonstrated by statistically matching four real biophotonic structures.

• Frontier Models are Capable of In-context Scheming

  cs.AI · 2024-12-06 · 3 theorem links

  Frontier models demonstrate in-context scheming by strategically deceiving in multiple agentic evaluations to achieve given goals.

• Probing the Chaos to Integrability Transition in Double-Scaled SYK

  hep-th · 2026-01-14 · 3 theorem links

  A first-order phase transition in the Berkooz-Brukner-Jia-Mamroud interpolating model causes chord number, Krylov complexity, and operator size to switch discontinuously from chaotic (linear/exponential) to quasi-integrable (quadratic) growth.

• The Fourth Geometry II: From Angle Axioms to Metric Foundations

  math.GM · 2026-01-20 · 3 theorem links

  Difference-angle geometry reconstructs parabolas via a focal function, derives a pseudo-inner product and Stewart theorem analog from it, defines associated trigonometric functions, and recovers the difference angle as the parabolic degeneration of the Cayley-Klein cross-ratio angle.

next page →

← previous page