The Archive

State-space architecture for facial expression recognition in video using Mamba backbone with linear-cost spatiotemporal scanning.

Comparative analysis of position bias across Transformer, state-space, and Mamba architectures in long-context QA tasks.

Evaluation protocol for personalized federated learning that reveals generalization gaps by testing global, rare-class, and unseen-class performance.

Formal analysis of offline agent verification limits via matched safe/harmful lifecycles, establishing error-abstention tradeoff for post-deployment auditing.

Forward-only training method for continual updates to language model n-gram memory without backpropagation or modifying pretrained weights.

Training-free method for coordinated parallel token commitment in few-step diffusion language models balancing reliability and joint consistency.

LLMs combined with constraint reasoning to generate contrastive explanations for MIP-based optimization decisions, addressing interpretability in DSS.

ProductWebBench: benchmark for frontend coding agents measuring both task completion and website continuity across 400 real-repo tasks.

PRISM: method for recognizing procedural mistakes in videos using HMM-grounded semantic state mapping and step-order validation.

Dataset of 121k ground-reason pairs from French Court of Cassation rulings (2016-2025) for legal domain analysis.

RULES: neuro-symbolic model for interpretable sequential event prediction via learned temporal rules.

AI agent via MCP for structural design translation from natural language to domain-specific 3D models using COMPAS API.

DRAFT: method reconstructing 3D room layouts from fMRI signals and camera trajectories without visual frames.

Active Preference Acquisition (APA) framework for web agents to identify high-value user preference questions via Prospective Personalization Value.

Multi-agent LLM consensus study: larger models absorb opinions of smaller models regardless of task design across seven model families.

DeepOpen: intent-classification system achieving 94.12% (BANKING77) and 98.02% (CLINC150) via SAM and ensemble methods.

A* search with domain-specific heuristics outperforms SAT/IP solvers on Flood-It puzzle optimization, offering generalizable algorithmic insights.

RIFT addresses decode-time KV cache staleness in speculative decoding via complementary compression training, improving long-context draft model efficiency.

Theoretical framework linking open-ended intelligence to recursive semantic self-extension and purpose transformation via Kantian/Gödelian foundations.

Reward model scores lack between-prompt comparability; human-referenced audits reveal pairwise evaluation blindness to shared prompt-level signal components.

The Eiliko robot toy. | Photo by Sean Hollister / The Verge Eiliko, a cute robot charm you can hang on your bag with built-in AI conversations, turned out to be a security nightmare. It told owners that "your secrets are safe with me," but hackers could easily read complete transcripts of everything we told the toy, and that was just the beginning. The Verge and security researcher Sammy Azdoufal spent over a month trying to get in contact with the company, including emails to Energize Lab CEO and founder Shaw Yeung. Nobody would answer our questions. Now, the company is finally breaking its ...

Transformers with CoT reasoning learn heavy-ball gradient acceleration without explicit supervision, exceeding teacher-student gradient descent imitation.

MQAScore evaluates text-to-music alignment via large audio language model on five atomic dimensions (genre, instrument, mood, vocal, usage).

General-purpose AI agents autonomously perform 3D medical image annotation via nnU-Net integration, requiring no model retraining.

Relearning phenomenon in knowledge-graph embeddings: forgotten facts recover during continued training due to shared entity representations.

Federated DP-PCA achieves linear communication cost for private covariance estimation in federated GLM optimization with central DP guarantees.

Dual Exam benchmark isolates multimodal agent action-selection from world-model state-evolution via fixed mechanisms across 15 domains and 1,488 test items.

SCOPE compiles natural-language policies into neurosymbolic programs for stable, efficient LLM agent governance over multi-turn trajectories.

Study reveals Theory of Mind benchmarks conflate explicit belief recognition with applied behavior prediction; Llama-3.1-8B shows 51.5% aggregate but uneven competencies.

Multi-directional Perturbation-Rewrite Search framework uses LLMs to automate heuristic design for NP-hard combinatorial optimization with factorized transformations.