Core topic
Moving AI agents from tools to accountable collaborators
Agent orchestration, collaboration, ownership, escalation, long-running work, and human oversight from an operational perspective.
A new evaluation setting that turns building a customer service agent into the task for coding agents. A close read of its realistic starting conditions, held-out user scoring, low pass rates, and what they mean.
Why curators who only read trajectories pollute memory, and how read-only environment probing that checks, scopes, and refreshes memories changes results on CLBench and APEX.
How UnitBoost replaces the generative manager with a unit map, constrained argmax assembly, and an explicit residual, with results and limits explained.
A close reading of a controlled web-task study on procedural memory reuse and interference, from the WebShop version history and a divergent pilot run to a 32-cell experiment and what its region of non-interference does and does not show.
LexAgentHallu traces legal agent failures across 3,414 cases and 27 subtypes, showing why final-answer scores hide broken reasoning along the way.
When should a diagnosis agent commit to a diagnosis and when should it step back, with selective error and autonomous coverage controlled together through the Cros stopping layer.
A reading of PRAGMA on how to evaluate assistants that live in long conversations, from the gap between factual recall and personal guidance to evidence alignment and the next memory design.
Beyond chat evals, a taxonomy-driven framework that automatically finds multi-step risks in agentic AI from the outside
Why ranking tools by similarity to the request leaves the final action stranded, and how a menu built as a prior over execution paths lifted online success on ToolBench.
A walkthrough of a human in the loop, multi-agent GeoAI framework that treats Arctic routing as a multi-criteria choice across operational, physical, ecological, and community concerns.
A reading of when in-context skill execution frays and when subagent delegation holds, through contracts, procedure, and coordination cost.
A walkthrough of RobustSGPO, which repairs the open choices in SGPO with explicit edit requests, patch construction and checks, and restarts from retained snapshots, and why that lifted completion and quality in the AgentX brainstorming flow.