From 74b1c1f20b57a901f5942784eb8ff4d0b39fb864 Mon Sep 17 00:00:00 2001 From: Richard Abrich Date: Tue, 3 Mar 2026 21:20:17 -0500 Subject: [PATCH] fix: add inline caveat to 100% accuracy claim in README table The table cell now reads "100% first-action (n=45, shared entry point)" instead of just "100% (validated, n=45)" so the limitation is visible even when the table is read without the surrounding paragraph. Co-Authored-By: Claude Opus 4.6 --- README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README.md b/README.md index be1a4103a..252b4f549 100644 --- a/README.md +++ b/README.md @@ -170,7 +170,7 @@ Zero-shot VLMs fail on GUI tasks not due to lack of capability, but due to **amb | | No Retrieval | With Retrieval | |---|---|---| -| **No Fine-tuning** | 46.7% (zero-shot baseline) | **100%** (validated, n=45) | +| **No Fine-tuning** | 46.7% (zero-shot baseline) | **100%** first-action (n=45, shared entry point) | | **Fine-tuning** | Standard SFT (baseline) | **Demo-conditioned FT** (planned) | The bottom-right cell is OpenAdapt's unique value: training models to **use** demonstrations they haven't seen before, combining retrieval with fine-tuning for maximum accuracy. Phase 2 (retrieval-only prompting) is validated; Phase 3 (demo-conditioned fine-tuning) is in progress.