The paper proposes Narration-of-Thought, a system prompting technique that structures chain-of-thought reasoning to improve ethical decision-making and reduce bias in large language models.
The Large Sparse Reconstruction Model demonstrates how expanding transformer context windows can improve the recovery of fine-grained textures in feed-forward 3D reconstruction.
A new multi-agent workflow utilizing large language models has been developed to autonomously formalize complex theorems in theoretical physics.
CausalDS is a new benchmark designed to evaluate the causal reasoning and data analysis capabilities of language model agents.
Researchers have introduced ProjAgent, a repository-level code generation system that retrieves functions based on procedural logic rather than simple lexical or semantic similarity.
Researchers propose a hierarchical framework that utilizes a centralized large language model for strategic planning to coordinate specialized reinforcement learning policies in multi-agent environments.
Researchers have introduced a proactive memory agent designed to mitigate behavioral state decay in long-horizon tasks by actively retrieving relevant context.
Researchers have introduced Concept-as-Tree, a synthetic data framework designed to improve the personalization capabilities of vision-language models by generating high-quality positive and negative samples.
Researchers developed liteOdyssey, a lightweight framework that guides general-purpose language models through diagnostic reasoning for rare diseases without altering model weights.
A new framework maps large language model personality traits within weight space using low-rank adapters to measure and control model personas.
The paper introduces SearchGen-20K and SearchGen-Bench to address the world-knowledge limitations of visual generators by integrating search capabilities.
VEGAS is a training-free evaluation metric that uses viewer gaze data to measure how well generated video captions align with human visual attention.
A new benchmark called PredicateLongBench has been introduced to systematically analyze the specific factors that make long-context tasks difficult for large language models.
The study presents a mechanistic taxonomy and diagnostic pipeline to track how failures like reward hacking and policy collapse emerge during reinforcement learning from human feedback.
Researchers have developed a content protection method that exploits context compression to prevent LLM-based web crawlers from scraping online information.
Researchers have introduced a three-stage curriculum learning framework designed to efficiently distill chain-of-thought reasoning from large language models into smaller student models.
CriterAlign is a criterion-centric evaluation framework designed to improve pairwise code preference judging by aligning rationales across specific criteria.
This study investigates how pre-trained code models internally represent programming types by probing their hidden states across different programming languages.
This study introduces a self-validating framework to analyze and mitigate safety hazards and hallucinations in LLM-assisted safety engineering tools.
Ant Group's embodied AI division has released new spatial perception and vision models designed to help robots accurately detect transparent and reflective objects like glass.
QbitAI has issued an invitation for readers to join their live broadcasts and explore the World Artificial Intelligence Conference 2026.
An Ivy League professor's decision to mandate an in-person exam due to concerns over AI-assisted cheating resulted in a fifty percent drop in student test scores.
The newly introduced IdeaGene-Bench evaluates the ability of artificial intelligence systems to trace scientific lineages and generate new ideas based on existing research.
ZendoWorld is a new interactive environment designed to test how well AI agents can perceive visual inputs, form hypotheses, and design experiments to infer hidden rules.