The Archive

MetaExp enables LLM agents to learn transferable experience across tasks via scalable cross-task environment feedback.

FlashAttention variant computes attention statistics without materializing full matrix, enabling efficient hallucination detection in frozen proxy models.

Survey of AI applications in humanitarian domains including weather forecasting, renewable energy, poverty prediction, and disaster response.

Kolmogorov-Arnold Networks exhibit exponential worst-case sample complexity on high-dimensional problems despite dimension-independent approximation theory.

Tool-augmented LLM framework separates variable extraction, formula execution, and formatting for reliable medical calculator scoring from patient notes.

Benchmark of 83 published works shows LLMs struggle to generate nonlinear narratives, preferring chronological event ordering over author-specific structures.

DuET dual-encoder jointly predicts piano score attributes and beat tracking via note-frame cross-attention and alternating autoregressive decoding.

Embedding-and-LLM workflow refines CASTON casting domain ontology by identifying subsumption errors in manually and automatically constructed hierarchies.

SpecAgent delegates tool-call sequences to smaller models with semantic verification, reversing speculative decoding to reduce strong model computation.

Knowledge Transfer Probing reveals uneven transfer of visual knowledge to language understanding in unified multimodal models through image-generation supervision.

TITAN framework uses causal intervention and sparse feature decomposition to identify and optimize translation circuits in multilingual LLMs for low-resource language pairs.

ELLSA 2 enables full-duplex dialogue with synchronized speech, gestures, and expressions using chunk-wise synthesis with distinct listening/speaking states.

X-Fly improves aerial vision-language navigation by executing full action chunks before replanning, reducing inference cost and addressing termination failures.

TIR-EC uses tool-integrated reinforcement learning to improve LLM performance on EHR-based emergency care prediction by addressing feature selection and risk judgment.

V-SHARE provides first complete convergence guarantees for advantage-based value-learning methods with cross-action information sharing in reinforcement learning.

SITAR is a situational distortion metric for ASR evaluation that captures semantic consequences of transcription errors beyond lexical differences.

MRTPO assigns multi-reward advantages to relevant tokens in language model training, improving credit assignment for decoupled reward signals.

HyCo hybrid speculative decoding optimizes mobile-cloud inference by assigning server multi-token prediction and SLM drafting to device with quantization.

PAV (Policy-Aligned Predictive Verifier) evaluates VLA action-chunk feasibility by anticipating prefix consequences rather than relying on generation stability signals.

Figma is also releasing its code layers feature in open beta.

Tiny Little Office web app suite uses URL fragments for shareable documents and provides llms.txt for agent-friendly URL construction.

The founders detailed how their startup is developing AI that can design chips, learn from each iteration, and improve subsequent designs, all with the goal of automating and accelerating the chip design process.

Framework for enterprises to measure AI ROI by quantifying costs, financial value, and building credible estimates.

Cohere acquires AISLE to strengthen sovereign AI security and resilience for enterprise and government customers.

Guidance for utility sector on AI deployment considerations including security and operational requirements.

Enterprise strategies for managing agent proliferation and leveraging controlled duplication to discover effective agentic solutions.

OpenAI and Sullivan & Cromwell launch Agreement Analyzer, an AI tool for contract review and issue identification.

Every day, conversations unfold across social media and websites around the world. A seemingly insignificant post—whether it involves AI-generated misinformation, hate speech, or a damaging rumor—can quickly snowball into a much larger public controversy. Gudea, a company that tracks how stories and rumors spread online that describes itself as the “storm tracker for the internet,” […]