All papers

RePara trains diffusion language models using group-relative reinforcement learning on short token blocks to improve math generation speed and accuracy.

The paper evaluates seven language models on estimating soil organic carbon concentrations, calculating carbon stock, and comparing agricultural paddock pairs.

SATResetCoP restarts connection-tableau search branches while preserving accumulated clauses in an incremental SAT solver to improve theorem proving efficiency.

TextualRL updates agent context prompts using groupwise trajectory comparisons rather than isolated feedback runs to improve policy adaptation without parameter tuning.

This paper audits a specific scalar feature scaling method to test whether claimed curvature benefits hold for domain generalization benchmarks.

The authors present a framework using level-wise multimodal memory to reconstruct coherent heading hierarchies across long, multi-page documents.

TrueMAD introduces one-to-one peer feedback between debate rounds to help language model agents resolve conflicting reasoning before final revision.

LiFUSE uses frequency-spatial self-supervised pretraining on non-contrast CT scans to improve representation learning for liver disease detection.

FALLNEST condenses large imbalanced wearable motion sensor datasets into small balanced synthetic sets for training fall detection models.

JevGPT combines fast local evidence assessments with deliberative planning to improve autonomous research agents operating under fixed retrieval budgets.

AnnalsChallenge provides 50 formalized mathematics theorem statements in Lean 4 drawn from papers published in the Annals of Mathematics.

AffectWorld benchmarks world models on predicting emotion-driven human actions and subsequent environment changes across annotated video samples.

SimInter simulates conversational recommendation users by conducting synthetic interviews between language models to enrich implicit behavioral profiles.

This paper examines using provenance reaction action graphs and machine learning to encode multi-step material synthesis pathways from experimental history.

ShoeFit presents a multi-view dataset and a dual-stream diffusion framework for consistent virtual footwear try-on synthesis.

AssetForge pairs reusable application assets with task rubrics to synthesize executable training tasks and evaluation checks for business automation agents.

The authors evaluate how 8-bit and 4-bit quantization methods affect model accuracy across Bavarian, Standard German, and English benchmarks.

Orchestra decomposes multimodal instructions into subtasks, tracks step progress, and maintains execution context to improve long-horizon robotic manipulation.

The authors formulate a meta-game framework to dynamically route and select language models during strategic multi-agent interactions.

The authors introduce an embodied reasoning foundation model that unifies perception, planning, and execution using task-conditioned adaptive adapters.

The authors present an active spatial search framework enabling vision-language agents to ground 3D objects in city-scale environments via interactive tool use.

The authors evaluate offline graph-modified embeddings for vector retrieval, improving multi-hop evidence recovery without requiring query-time graph access.

The authors trace internal representations across prompt optimization techniques to show how discrete prompts, prompt tuning, and prefixes alter model processing.

The authors propose a resource-efficient two-stage approach for estimating 3D road user shapes and orientations from roadside cameras and radar.

The authors introduce a 2.5-billion-parameter bilingual Tamil-English language model trained from scratch with tool-calling capabilities.

The authors introduce a topology-aware molecular representation language that enables fragment-level representation learning and precise structural edits.

The authors formulate a stochastic model to analyze the timing distributions and detection limits of potential recursive artificial intelligence explosions.

The authors propose an embodied world model that unifies RGB video, metric depth, and robot actions into a shared dynamic representation.

The authors introduce a benchmark to evaluate autonomous scientific discovery agents making iterative experimental decisions under fixed budgets.

The authors present an energy-based prior for cellular perturbation models to better distinguish specific therapeutic signals from general biological background noise.

The authors propose a memory-driven vision-language navigation system using temporal compression and geometric-semantic representations to support long-horizon trajectory tasks.

The authors present a training framework and dataset that enables terminal coding agents to self-improve by recovering from execution errors.

The authors analyze task-evoked eye movements in adolescents to identify objective behavioral markers corresponding to depression severity scores.

The authors propose evaluating autoformalization faithfulness by aligning phrase-level correspondences between informal mathematics and formal code under contextual interpretations.

The authors introduce a benchmark evaluating egocentric video understanding structured around an interactive loop of user perception, action, and world feedback.

The authors introduce a challenge benchmark to develop foundation models for building HVAC energy optimization that transfer across different physical structures.

The paper analyzes the geometric distribution of language model output logits, identifying a consistent log-linear relationship that emerges during pretraining.

The paper presents an extended behavior tree framework that automates modular robot task planning and runtime replanning for off-site construction fabrication.

The paper defines an algebraic symmetry framework for characterizing structural agent heterogeneity and policy representations in multi-agent systems.

The paper presents a test-time search method that perturbs input conditions to generate diverse reasoning paths in recursive reasoning models.

The paper analyzes pathwise distance consistency in visual world models to avoid candidate order reversals during latent short-horizon planning.

The paper introduces an agent-based approach that adapts search spaces to optimize post-training recipes and configurations for pre-trained language models.

The paper analyzes implicit morals in news coverage of artificial intelligence across four global regions, identifying distinct regional variations in narrative caution.

The paper theoretically analyzes multi-objective best-arm identification to determine when multiple criteria reduce exploration requirements in multi-armed bandit settings.

The paper presents a multimodal emotion recognition framework that integrates Euclidean and hyperbolic geometry to capture hierarchical emotion structures from text and speech.

The paper proposes a continual off-policy reinforcement learning framework that reuses actor and critic networks across sequential robotic manipulation tasks.

The paper presents a generative audio method that controls Foley sound synthesis using physical interaction, material, and geometry parameters.

The paper examines projector training in multimodal models and proposes a training-free initialization method using summary statistics instead of paired image-caption data.

The paper adapts a causal language model using compressed 3D geometric tokens to assemble multi-part objects in a single forward pass.

The paper introduces a benchmark to evaluate how typed decision models handle theory-of-mind reasoning tasks compared to standard generative language models.

The paper evaluates whether self-distillation using privileged information improves reasoning capabilities or reflects overfitting to training tasks.

This paper proposes a system where language model agents write code to create and refine tools directly over raw knowledge-graph files.

The authors compare information precision and preference reweighting mechanisms in active-inference agents managing competing survival needs.

The paper characterizes optimal falsification-proof non-market allocation mechanisms when individuals can strategically manipulate their observable characteristics.

The authors present an approach using local language models to automatically generate conceptual architectural models and reports for digital twins.

This paper introduces a metric to estimate the compatibility of pretrained models before running costly alignment or checkpoint merging.

The authors reformulate a recurrent video prediction architecture using gated recurrent units to reduce parameters while maintaining spatiotemporal modeling.

This paper applies recursive training and discrete deduction procedures to message-passing neural networks to solve constraint-reasoning tasks with few parameters.

The authors introduce a benchmark to evaluate how audio models update, retain, or reject recurring acoustic sources in long streams.

This technical report presents an updated summary-token attention architecture to optimize cache compression and computation across model layers.