6239 papers
The paper introduces CRAFT, a low-cost, open-source, tendon-driven anthropomorphic hand that combines rigid links with soft joints and rolling-contact surfaces to achieve robust, repeatable, and versatile contact-rich manipulation.
This paper demonstrates that increasing the sophistication and diversity of AI agents can paradoxically worsen collective outcomes and system overload when resources are scarce, with the overall impact depending entirely on the capacity-to-population ratio rather than the agents' inherent intelligence.
This paper introduces TopoBench, a benchmark for evaluating large language models on topological grid puzzles, revealing that their poor performance stems primarily from difficulties in extracting and maintaining spatial constraints rather than inherent reasoning limitations.
This paper introduces a cost-effective, automated recipe combining generic prompts, hierarchical verification, and iterative agent-assisted repair to translate complex reinforcement learning environments into high-performance implementations with zero sim-to-sim gap, achieving massive throughput gains (up to 22,320x) across diverse use cases including game emulation, physics simulation, and card game engines.
FlashMotion is a novel training framework that enables high-quality, few-step trajectory-controllable video generation by combining a pre-trained trajectory adapter with a hybrid diffusion-adversarial finetuning strategy, while introducing the FlashBench benchmark to evaluate performance across varying object counts.
This paper establishes compute-optimal scaling laws for on-policy LLM reinforcement learning by demonstrating that the ideal number of parallel rollouts per problem increases predictably with the compute budget before saturating, driven by solution sharpening on easy tasks and coverage expansion on hard ones, while providing practical allocation rules for batch size and update steps to maximize training efficiency.
GlyphBanana introduces a training-free, agentic workflow that integrates auxiliary tools to inject glyph templates into latent spaces and attention maps, enabling various Text-to-Image models to achieve superior precision in rendering complex text and mathematical formulas without requiring retraining.
This paper provides a theoretical framework for quantifying forgetting in continual post-training of generative models by analyzing how divergence objectives (forward vs. reverse KL), geometric mode overlap, and replay strategies interact to cause either mass forgetting or controlled component drift.
BehaviorVLM is a unified, finetuning-free framework that leverages pretrained Vision-Language Models with explicit reasoning steps to achieve scalable, label-light pose estimation and behavioral understanding for freely moving animals without relying on extensive human annotation.
This paper introduces the MADQA benchmark and a novel accuracy-effort evaluation protocol to demonstrate that while multimodal agents can match human accuracy on document-based tasks, they rely on inefficient brute-force search rather than genuine strategic reasoning, failing to close the performance gap to oracle levels.
This paper introduces Proof-Carrying Materials (PCM), a rigorous framework combining adversarial falsification, statistical refinement, and formal Lean 4 certification to overcome the high failure rates of single machine-learned interatomic potentials, thereby significantly improving the reliability and discovery yield of high-throughput materials screening.
This paper presents a practical, polynomial-time compilation method that translates temporal planning with durative actions into the PDDL+ language, fully capturing the semantics while retaining plan length up to a constant factor and demonstrating effectiveness on hard temporal numeric problems.
This paper introduces WORKSWORLD, a new domain for numeric planners that automates the joint planning and scheduling of distributed data pipelines by allowing users to define high-level goals without specifying the entire workflow graph, demonstrating the ability to solve complex multi-site problems on commodity hardware.
This paper proposes RDNet, a salient object detection network for optical remote sensing images that leverages a SwinTransformer backbone and three novel modules—Dynamic Adaptive Detail-aware, Frequency-matching Context Enhancement, and Region Proportion-aware Localization—to overcome challenges related to scale variations and global context modeling, thereby achieving superior detection performance compared to state-of-the-art methods.
This paper presents Portfolio-CEGAR-SEQ, a parallelized algorithm that leverages modern multi-core CPUs and a portfolio of diverse object arrangement strategies to outperform the original CEGAR-SEQ method in solving the combinatorial challenges of object arrangement and scheduling for sequential 3D printing, often resulting in more efficient use of printing plates.
The paper introduces Idea-Catalyst, a novel LLM-driven framework that enhances scientific creativity by systematically decomposing research goals into domain-agnostic problems to retrieve and synthesize interdisciplinary insights, thereby improving the novelty and insightfulness of brainstorming without premature solution anchoring.
This paper argues that large pretrained models contain a dense distribution of task-specific experts near their initial weights, enabling a simple, parallel post-training method that samples random perturbations and ensembles the best performers to achieve competitive results with standard optimization techniques like PPO and GRPO.
Drawing from Perplexity's operational experience with general-purpose agentic systems, this paper outlines the unique security failure modes introduced by AI agents, maps their primary attack surfaces, proposes a layered defense strategy, and identifies critical research gaps and standards needed to secure multi-agent systems in alignment with NIST risk management principles.
This paper proposes an incremental neural network verification technique that reuses learned conflicts across related queries via a SAT solver to prune infeasible search spaces early, achieving speedups of up to 1.9x over non-incremental baselines.
This paper introduces the separable neural architecture (SNA) as a domain-agnostic primitive that unifies predictive and generative intelligence across physics, language, and perception by formalizing a structural inductive bias that factorizes high-dimensional mappings into low-arity components, thereby enabling effective modeling of both chaotic continuous systems and discrete sequences.