xuan/memory/MEMORY.md

- [Qwen3-4B CPT Experiment](qwen3-4b-cpt-experiment.md) — 20B token CPT on 8xH100, completed 2026-05-22, final ckpt at global_step_19074
- [Environment Setup](env-setup.md) — conda env `ttt`, VeOmni, dependencies, lr_decay_ratio gotcha
- [N4 Transfer Info](n4-transfer.md) — sigma@135.232.252.197:50000, SSH key not configured yet
- [Qwen3-4B-NHS Model](qwen3-4b-nhs-model.md) — YoungXuan/Qwen3-4B-NHS; In-Place-TTT inference silently DISABLES its TTT (zero conv)
- [NIAH Eval Results](niah-eval-results.md) — Single-needle NIAH saturated (Base 9.96, NHS 10.00), not useful for paper
- [BABILong Eval Results](babilong-eval-results.md) — earlier NHS "boost" INVALID (TTT was a no-op); Base-v2 is the valid baseline
- [TTT Inference Findings](ttt-inference-findings.md) — faithful TTT replay collapses; NS conditions update DIRECTION; needs tuned-down length-aware lr
- [Qwen3-0.6B TTT Experiment](qwen3-0.6b-ttt-experiment.md) — In-Place-TTT 1k-step train; faithfully reproduces ttt_lr=3 grad explosion (in progress)
- [Inference Always cf-patch](inference-always-cfpatch.md) — RULE: inference always uses cf-patch fast-weight update; report ONLY cf-patch score, never the "vanilla"/no-TTT column
- [TTT Train Recipe Findings](ttt-train-recipe-findings.md) — RULE: ttt_norm_preserve=TRUE is mandatory (+20-30pt RULER, fixes loss-blowup); inner η insensitive plateau w/ norm=true; test η~3 (paper's value)