EASE-TTT: Evidence-Aligned Selective Test-Time Training for Long-Context Question Answering

In three linesEASE-TTT combines evidence retrieval and test-time adaptation to improve long-context QA. The method converts selected evidence chunks into soft attention supervision targets, guiding query-side parameter adaptation. Tested on 6 LongBench tasks with 3 small decoder-only models, it outperforms retrieval-only baselines and qTTT.

Summary generated by Claude — human-verified