Guy

The paper introduces LACUNA, a novel programming model that allows LLM agents to write code that shapes the runtime environment while maintaining strong type-checking safety guarantees.

The paper introduces a distributional framework using Wasserstein distance to unify the semantic comparison of sparse autoencoder features across different layers and to automatically compress large feature circuits into interpretable supernodes.

MolLingo is a multi-agent system that significantly improves automated molecular design by integrating domain-specific chemical reasoning and structural context into LLMs, outperforming state-of-the-art models on multiple benchmarks.

FPMoE introduces a sparse Mixture-of-Experts (MoE) architecture to improve functional code generation across multiple functional programming languages, achieving state-of-the-art performance with fewer parameters.

This case study demonstrates that in complex scientific software development, human domain expertise and careful supervision are more critical to ensuring the trustworthiness of AI-generated code than the raw capability of the AI model itself.

The paper introduces partial multi-neuron relaxation, a novel verification technique that selectively computes tight linear bounds for a small subset of neurons to improve the efficiency and tightness of neural network safety verification.

The paper introduces Croissant Tasks, a declarative metadata format designed to achieve conceptual reproducibility in machine learning by abstracting problem specifications from brittle implementation details.

The paper proposes a novel decision mechanism that decouples the decision to issue an alert from the raw likelihood of conversational derailment, significantly reducing false positives by simulating potential recovery paths.

This paper refines word-based grammatical error annotation for L2 Korean by adapting existing resources to better reflect Korean morphology and error types, improving the evaluation of Korean Grammatical Error Correction (K-GEC) systems.

This paper introduces a welfare-centric framework for designing institutional incentives, showing that optimizing for total social welfare often requires different incentive levels than those optimized for cost or cooperation frequency.

CodeCytos is a novel coding-based reasoning agent framework that enables dynamic, programmable interaction with spatial molecular imaging data, significantly improving the automation and customization of complex tissue analysis.

The paper introduces NICE, a declarative framework that uses NixOS to build and automatically validate reproducible environments for demonstrating software vulnerabilities (CVEs), thereby improving the reliability and shareability of security research.

The paper introduces Curvature-Conditioned Query (CCQ), a novel read-time contraction mechanism that improves linear attention's performance on long-context and retrieval tasks by incorporating the geometry of the softmax function.

The paper introduces MIDI, a novel multilingual dataset that embeds idioms in realistic sentence and conversational contexts across diverse resource levels, revealing that idiom comprehension is significantly harder in low-resource languages and that literal interpretations pose a greater challenge than figurative ones.

The paper proposes TriAlign, a novel multi-agent reinforcement learning framework that achieves universal truth consistency across social groups in personalized LLMs while maintaining high accuracy and personalization.

The paper introduces ProbMoE, a probabilistic routing framework that tackles the non-differentiability of top-$k$ routing in Mixture-of-Experts (MoE) models, achieving strong performance with improved expert utilization.

This paper optimizes the decoding of Hamming Quasi-Cyclic (HQC) codes for post-quantum cryptography on NPU-integrated mobile devices by redesigning the kernels to leverage the Hexagon Vector eXtensions (HVX), achieving significant reductions in latency and energy consumption.

The paper proposes a novel probabilistic globally constrained decoding (P-GCD) method that efficiently constructs proposals for locally constrained decoding, significantly improving convergence speed and performance compared to existing approaches.

This paper optimizes the decoding of Hamming Quasi-Cyclic (HQC) codes for post-quantum cryptography on NPU-integrated mobile devices by redesigning the core kernels to leverage the Hexagon Vector eXtensions (HVX) backend.

The paper presents a new worst-case to average-case reduction for the Learning Parity with Noise (LPN) problem, achieving hardness for inverse-polynomial noise rates previously unattainable.