The Morning
From the arXiv
ACEA: An Adversarial Co-Evolution Arena for Head-to-Head Red-Team and Blue-Team LLM Testing
CEA is a novel platform that addresses the isolated development of LLM red-team attacks and blue-team defenses. It facilitates head-to-head adversarial testing by connecting pluggable attack and defense modules to a shared LLM, using an LLM judge for scoring. ACEA's key contributions include a model-agnostic arena with a standardized adapter protocol (ASAP) and an evaluation methodology designed for adversarial rounds with verifiable ground truth.
AMA: Adaptive Memory via Multi-Agent Collaboration
AMA addresses limitations in LLM agent memory by introducing a novel framework that uses multi-agent collaboration. Its core method is a hierarchical memory design that dynamically adjusts retrieval granularity based on task complexity. This allows for more ef…
Breaking Planner Integrity Boundary: Enviroment State-Text Injection Attack on LLM-Driven Embodied Agents
This paper introduces the Environment State-Text Injection (ESTI) attack, a novel method that targets LLM-driven embodied agents by manipulating the text describing their environment. Unlike previous attacks, ESTI directly injects deceptive state information, …
Closing the Consistency Gap: Self-Evolving Agents That Learn to Stay on Course
This paper addresses the "consistency gap" in LLM agents, where their average accuracy hides significant unreliability across multiple runs. Their core method is a "self-evolving agent framework" that identifies and learns from inconsistent steps. By convertin…
Compressing Sequences in the Latent Embedding Space: $K$-Token Merging for Large Language Models
This paper introduces K-Token Merging, a novel method for compressing long prompts in LLMs by merging contiguous token embeddings in the latent space. This approach, unlike previous token-space methods, reduces computational and memory costs by creating a shor…
Does Deeper Reasoning Compromise Alignment? Revealing and Mitigating of Alignment Collapse in Large Reasoning Models
This paper reveals that deeper reasoning in Large Reasoning Models (LRMs) can paradoxically lead to "alignment collapse," degrading their safety. They introduce the Alignment Loss …
Graph-Based Personalized Memory for LLM Agents: Representation, Evolution, Retrieval, and Evaluation
This paper proposes a graph-based framework for LLM agents to build personalized, long-term memory. It structures user information and interactions as a graph, enabling the agent t…
Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models
This survey explores "reasoning economy" in LLMs, aiming to balance accuracy with computational cost. It analyzes the causes of inefficient reasoning and proposes solutions for bot…
Inference-Time Nash Alignment
This paper introduces inference-time alignment methods, Best-of-Nash (BoN) and Nash Mirror Descent (NMD), that can align language models with general preferences without updating m…
Inferring the Unspoken: Aligning Embodied Agents with Implicit Preferences
This paper introduces a method called Inferring the Unspoken (InTU) that enables embodied agents to learn and apply implicit user preferences from observed behaviors. The core cont…
The Town Square
Code suggests Apple's Siri could be replaced by third-party AI models like ChatGPT or Claude, offering users more advanced conversational abilities.
Workshops
This repository provides a fast, battle-tested hybrid code review tool for Alibaba's scale, combining deterministic pipelines with LLM agents for precise, line-level comments and multi-language security rule checks.
YuE2 is a frontier music generation system that leverages symbolic planning, zero-shot cover creation, and agentic editing for advanced musical composition.
