The Morning
From the arXiv
ARISE: Adapting to Evolving Capability Gaps in Agentic Reinforcement Learning
RISE addresses the challenge of evolving capability gaps in long-horizon reinforcement learning agents. Its core method, Adaptive Rubric-Skill Co-Evolution (ARISE), continuously adapts evaluation criteria and exploration strategies based on agent performance. This allows the agent to focus learning on emerging weaknesses and make progress even with sparse task-level feedback.
Continuous Context Management
This paper introduces Continuous Context Management (CCM) for LLM agents, which compacts interaction history at every turn instead of waiting for a threshold. This prevents the active prompt from growing excessively, significantly reducing input usage and prom…
KV-streams for Efficient Compaction in Agentic Reinforcement Learning
This paper introduces KV-streams, a novel method to improve the efficiency of context compaction in agentic reinforcement learning. Instead of repeatedly refilling the LLM's context, KV-streams stream the KV cache forward, significantly boosting training throu…
SEABench: Benchmarking Endogenous Misalignment In Self-Evolving Agents
This paper introduces SEABench, a novel benchmark designed to evaluate "endogenous misalignment" in self-evolving AI agents. The core method involves creating longitudinal task sequences in a personal assistant environment to observe how agents' self-modificat…
TokenCast: Forecasting Token Consumption During LLM Agent Execution
TokenCast addresses the unpredictable token consumption of LLM agents by learning a composable cost representation for execution segments. This method estimates future token usage by accounting for both segment-specific consumption and the context growth intro…
Rethinking Personalized Generation: Test-Time Alignment via Factorized Ranking Models
This paper proposes a test-time alignment method for personalized generation with LLMs. It argues that personalization is a candidate matching problem, not a generator limitation, …
ScAn-Bench: Evaluating Scaling Analysis Methodology
This paper introduces ScAn-Bench, a novel benchmark for evaluating methodologies used to derive scaling laws and optimal prescriptions for large foundation models. The authors syst…
AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation
This paper introduces AutoRef, a method to optimize the "harness" in agentic multi-reference image generation. AutoRef automatically tunes how reference images are processed and ho…
BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation
BaRe-Mem is a Bayesian memory system that estimates advisor reliability in multi-agent systems by analyzing the central model's beliefs and past interactions. This allows it to dyn…
Behavioral Foundation Models for Quality Diversity
This paper introduces BFM-QD, a framework that leverages the latent space of Behavioral Foundation Models (BFMs) for Quality-Diversity (QD) search. Instead of searching in high-dim…
The Town Square
Cal Newport argues that the rapid, unchecked development of AI labs necessitates external investigation to understand their societal impact and potential risks.
Workshops
Paperclip is an open-source application designed to help teams manage and orchestrate AI agents effectively in a work environment.
Hindsight is an AI agent memory system that learns and adapts over time, enabling agents to recall and utilize past experiences to improve future decision-making and performance.
