From da75668d6b31860c9362e93bab425e947164a7a9 Mon Sep 17 00:00:00 2001 From: wassname <1103714+wassname@users.noreply.github.com> Date: Tue, 28 Apr 2026 09:09:52 +0800 Subject: [PATCH] move RESEARCH_JOURNAL and fork_plan under docs/ Working notes belong with the rest of the docs. Updated relative links in docs/hypothesis_ablation_catalog.md from ../fork_plan.md to fork_plan.md since both files now live in docs/. Co-Authored-By: Claude Opus 4.7 --- RESEARCH_JOURNAL.md => docs/RESEARCH_JOURNAL.md | 0 fork_plan.md => docs/fork_plan.md | 0 docs/hypothesis_ablation_catalog.md | 10 +++++----- 3 files changed, 5 insertions(+), 5 deletions(-) rename RESEARCH_JOURNAL.md => docs/RESEARCH_JOURNAL.md (100%) rename fork_plan.md => docs/fork_plan.md (100%) diff --git a/RESEARCH_JOURNAL.md b/docs/RESEARCH_JOURNAL.md similarity index 100% rename from RESEARCH_JOURNAL.md rename to docs/RESEARCH_JOURNAL.md diff --git a/fork_plan.md b/docs/fork_plan.md similarity index 100% rename from fork_plan.md rename to docs/fork_plan.md diff --git a/docs/hypothesis_ablation_catalog.md b/docs/hypothesis_ablation_catalog.md index 2022d7e..9c1d45d 100644 --- a/docs/hypothesis_ablation_catalog.md +++ b/docs/hypothesis_ablation_catalog.md @@ -14,11 +14,11 @@ Core fork-plan mapping: | fork-plan experiment | Fits what | Does not fit what | |---|---|---| -| [cross-adapter causal `dW` basis ablation](../fork_plan.md) | learned `dW` SVD bases, shared adapter bases, per-adapter top/tail bases | pure activation bases unless first converted into a weight projection of the trained `dW` | -| [layer/module causal ablation of trained `dW`](../fork_plan.md) | layer slices, residual writers, attention output, MLP down, read/write module families | candidate bases that mix all layers without layer labels | -| [adapter-parameterization causal ablation of trained `dW`](../fork_plan.md) | LoRA rank components, PiSSA/DeLoRA S-space crops, DoRA magnitude vs direction, OFT rotations, IA3 gates | post-hoc activation PCA unless used only as an evaluation target | -| [activation-steering baseline](../fork_plan.md) | TaskDiff/RepE directions built without trained `dW`, selected on held-out validation rows | trained `dW` components, `act_oracle`, `TaskDiff_lora_fit` | -| [synthetic `dW'` baseline](../fork_plan.md) | pretrained read/write bases with signed coefficients from contrast activations | causal claims about the already trained adapter | +| [cross-adapter causal `dW` basis ablation](fork_plan.md) | learned `dW` SVD bases, shared adapter bases, per-adapter top/tail bases | pure activation bases unless first converted into a weight projection of the trained `dW` | +| [layer/module causal ablation of trained `dW`](fork_plan.md) | layer slices, residual writers, attention output, MLP down, read/write module families | candidate bases that mix all layers without layer labels | +| [adapter-parameterization causal ablation of trained `dW`](fork_plan.md) | LoRA rank components, PiSSA/DeLoRA S-space crops, DoRA magnitude vs direction, OFT rotations, IA3 gates | post-hoc activation PCA unless used only as an evaluation target | +| [activation-steering baseline](fork_plan.md) | TaskDiff/RepE directions built without trained `dW`, selected on held-out validation rows | trained `dW` components, `act_oracle`, `TaskDiff_lora_fit` | +| [synthetic `dW'` baseline](fork_plan.md) | pretrained read/write bases with signed coefficients from contrast activations | causal claims about the already trained adapter | | new causal test | nonlinear clusters, token-conditional attention routing, concept-space probes, DAS/SAE features | simple keep/drop of a fixed linear `dW` basis unless linearized first | ## Source provenance