173 paper takes since 2026-08-19
IAR post-trains LLMs in three stages—inject, align, recover—to internalize document knowledge for retrieval-free QA without sacrificing general capability.
Hierarchical self-modification evolves execution harnesses around frozen LLMs, yielding strong gains on moderate tasks but bounded by feedback quality.
A Japanese long-form video benchmark for evaluating narrative evolution and cultural reasoning, built with native-speaker verification.
A benchmark and dataset for evaluating whether generated videos achieve intended outcomes with semantic grounding.
SemaPLC gates PLC code acceptance on external compilation plus live runtime execution, lifting verified pass rates above baseline methods.
Co-RL trains multi-agent RL with peer-derived rewards to produce unsupervised reasoning across text and vision tasks without ground-truth labels.
Every pick in the Wire gets the same treatment: read, verified, and given a written verdict.
A self-play RL framework where a language model designs adaptive executable training environments and learns to solve them via regret-based targeting.
AdaPop adapts gradient pressure by fact popularity and automates the forget-retain balance to reduce leakage of unlearned content.
Centered residual signatures in weight space distinguish genuine fine-tune lineage from distilled or independent models — no training data needed.
Looped LLMs use recurrent passes to handle multi-step tool chains, with adaptive inference letting you trade compute for accuracy at runtime.
Action-conditioned objectives improve latent geometry for Euclidean-cost MPC by better aligning world-model representations with decision metrics.
Separating thematic seeds from genre-form controls yields diverse training data across 13 genres and measurably lifts LLM creative writing quality.
Every pick in the Wire gets the same treatment: read, verified, and given a written verdict.
SoftVTBench pairs synchronized vision and tactile data with a deformation-aware benchmark for evaluating robot manipulation of soft objects.
A study finding that skills improve LLM agents by stabilizing execution via procedural anchoring, not by supplying missing knowledge.
Evolution-strategy fine-tuning for long-horizon LLM agents with full-parameter updates and minimal GPU requirements, using trajectory-level reward weighting.
TAMP-Nav aligns vision-language models with 2D visual prompting and compressed selective memory to improve embodied navigation efficiency.
V-RAE builds semantically organized video latents from frozen vision representations to improve generation quality, convergence, and predictive modeling.
A framework for reproducible reinforcement learning across arbitrary agent harnesses, targeting improved coding-agent performance with minimal data and compute.
Every pick in the Wire gets the same treatment: read, verified, and given a written verdict.
A diffusion bridge method that translates low-resolution edits into high-resolution outputs while preserving source image details via sparse attention.
A black-box, zero-shot prompt-level defense that uses distribution-guided suffix expansion and contrastive scoring to reduce harmful image generation.
LEGO-RL pipes coding-agent harnesses directly into policy-gradient training and shows sparse MoE gains across multiple harnesses.
Multi-byte prediction cuts inference time in byte-level hierarchical LMs using variable-length windows and causal masking, with minimal quality loss.
A benchmark framework that uses hierarchical sub-agents to decompose world-model evaluations into verifiable reasoning chains with transparent evidence.
A recurrent Large Discovery Model pairs generative proposals with a Bayesian non-parametric reward surrogate, tested across molecules, proteins, and programs.