SOURCE-LINKED INTELLIGENCE

MemoryWalker: Stop Training Agents on Contexts They Never Saw

arXiv · AI, language, vision and robotics · article · Sep 1, 2026 · UTC

Production agent harnesses such as Claude Code and Qwen-Agent compress context during rollout, but training under compression creates a conditioning problem: every eviction branches the effective history, so the learning object is a tree rather than a sequence. Existing linearizations either retain the rightmost path, causing time-travel leakage, or replay a depth-first traversal, causing train-inference mismatch. We introduce two exact, gradient-equivalent corrections: LogitTree, a segmented K-forward traversal, and a packed 4D attention mask. LogitTree requires K+1 backward passes; the 4D ma

Read original source ↗ Open in workspace

recordType

paper

region

Global