All news

Assesses claims and costs of structural attacks on Classic McEliece.

Compares pipelining and combining for convergecast aggregation when values exceed one message.

AgenticEvolve uses persistent agents that design and run experiments inside evolutionary program search.

OLMo-3 checkpoints show little evidence of instrumental convergence in RLVR.

Typed template fuzzing for programs with complex structured inputs.

Agentic audit of Slovak FineWeb2 finds language-specific pipeline failures.

Active preference acquisition for long-term personalized web agents.

Co-adaptive hierarchical language modeling with continuous semantic compression.

Proposes sensorimotor alignment as a cheap offline proxy for robot policy success.

Isolates context in chain of thought to stop attention leakage.

Inside-JEPA adds relational grounding to latent world models.

Writes continual updates forward-only into an LM native n-gram memory table.

Diffusion LMs find natural-looking black-box attack prompts that trigger overthinking.

User simulators that stay on task and sound natural for multi-turn evaluation.

Analyzes what memory is actually worth to language agents.

Optimal learning of discrete distributions under person-level differential privacy.

Toolkit to audit differentially private graph releases.

Finds token collapse in graph soft prompting for LLMs.

Federated DP optimization via low-rank preconditioning.

Step-sequence reasoning to recognize mistakes in procedural videos.

Harness-evolving recursive distillation for continual learning.

Traces membership imprints in on-policy self-distillation.

Pre-history memory for humanoid loco-manipulation.

Advantage-based rule for when on-policy distillation should follow the teacher.

Compares search space complexity for end-to-end ASR.

DIAL distills controllable instruction-adherence levels into LLMs.

Change-aware multi-view commit message generation.

Harmful responses work as templates to jailbreak diffusion LMs.

How depth growth and looping shape in-context learning.

Shows much apparent rare-word TTS error is really ASR verifier error.