The Morning
From the arXiv
ALIVE: Awakening LLM Reasoning via Adversarial Learning and Instructive Verbal Evaluation
LIVE addresses the "reward bottleneck" in LLM reasoning by moving beyond costly scalar rewards. Its core method, Adversarial Learning with Instructive Verbal Evaluation, unifies problem posing, solving, and judging within a single policy. This allows LLMs to internalize reasoning logic directly from raw text through adversarial training and verbal feedback, fostering a deeper, self-contained understanding of correctness.
Co-ReAct: Rubrics as Step-Level Collaborators for ReAct Agents
Co-ReAct introduces a novel framework where rubrics act as step-level collaborators for ReAct agents. Instead of just evaluating final outputs, Co-ReAct integrates rubrics directly into the agent's decision-making process during inference, guiding each reasoni…
DiLaDiff: Distilled Latent-Augmented Diffusion for Language Modeling
DiLaDiff addresses the token correlation issue in diffusion language models by introducing a continuous latent space. This latent space, learned via an auto-encoder and a latent diffusion model, captures semantic information. A consistency model then distills …
Efficient and Transferable Agentic Knowledge Graph RAG via Reinforcement Learning
This paper introduces KG-R1, an agentic framework that uses reinforcement learning to optimize knowledge-graph retrieval-augmented generation (KG-RAG). Instead of fixed multi-module pipelines, a single agent learns to interact with knowledge graphs, retrieving…
Foundation Protocol: A Coordination Layer for Agentic Society
The Foundation Protocol (FP) introduces a graph-based coordination layer for agentic societies, unifying diverse entities like agents, humans, and institutions. Its core method is to provide a flexible framework for multi-party organization, event-based collab…
GENSTRAT: Toward a Science of Strategic Reasoning in Large Language Models
This paper introduces GENSTRAT, a novel benchmark for evaluating strategic reasoning in LLMs. Its core method involves procedurally generating a diverse distribution of imperfect-i…
Goal-Conditioned Agents that Learn Everything All at Once
This paper introduces Learning Everything All at Once (LEO), a method for goal-conditioned reinforcement learning. LEO efficiently learns from all transitions by jointly outputting…
It's the humans, not the data: Geopolitical bias in LLMs originates in post-training, amplified by the language of the prompt
This paper demonstrates that geopolitical bias in LLMs primarily emerges during post-training, not pre-training, contradicting common assumptions. The researchers found that models…
MemAudit: Post-hoc Auditing of Poisoned Agent Memory via Causal Attribution and Structural Anomaly Detection
MemAudit is a post-hoc framework for auditing poisoned memory in LLM agents. It uses causal attribution to identify memories that causally influence harmful outputs and structural …
Model Spec Midtraining: Improving How Alignment Training Generalizes
This paper introduces Model Spec Midtraining (MSM) to improve how language models generalize during alignment training. MSM involves training models on synthetic documents discussi…
The Town Square
Frontier LLMs exhibit significant disagreement when performing real-world fact-checks, highlighting challenges in their reliability for factual verification.
Workshops
This Python tool, microsoft/markitdown, converts various file formats, including Office documents, into Markdown, facilitating content repurposing and integration.
This plugin enables Claude Code, Codex, Cursor, and other AI coding assistants to interact with and leverage the Compound Finance protocol.
