Update README.md

README.md

@@ -198,10 +198,8 @@ Two modalities are provided: **amino acids** (per-protein SwissProt sequences) a
 
 ## ⭐ Which version to use
 
-| Path | Status | Use it? |
-|------|--------|---------|
-| **`EC_v2/`** | corrected, leakage-free, matches the current preprint | ✅ **Yes — all new work** |
-| `EC_v1/`     | original release (v1 preprint + parallel works); legacy | only to reproduce already-published v1 results |
+
+**`EC_v2/`** should be used for all new work, as it corrects important bugs in v1 data generation pipeline. `EC_v1/` is maintained for reproducibility of existing citing work and should only be used to reproduce already-published v1 results
 
 ## Structure
 
@@ -245,21 +243,6 @@ labels — they are **not** expanded into separate rows.
 237,421 proteins · 6,393 EC labels (1,321 compound) · 65,996 UniRef50 clusters.
 Sequences from UniProt release **2025_02**.
 
-## How GRIMM-EC v2 is built (and how it differs from v1)
-
-v2 reuses v1's UniRef50 cluster assignments but regenerates the splits to match the
-documented method:
-
-- **Per-protein SwissProt sequences** (release 2025_02) — v1's AA data instead held
-  the UniRef50 *representative* sequence.
-- **Low-support labels split by UniRef50 cluster** — labels with 1–2 clusters are
-  partitioned by whole cluster (2 clusters → 1 train / 1 test1; 1 cluster → orphan,
-  ~80% train / ~20% test2 across folds), not by individual sequence as in v1.
-- **Independent, shuffled folds**; **seeded** for reproducibility.
-- **`test2` is held-out only** (true open-set) — v1 inadvertently also wrote the
-  held-out orphans into train.
-- EC labels normalized (stray whitespace stripped).
-
 **Verified for v2 (all 5 folds, both modalities):** 0 `(sequence, EC)` overlap between
 `train` and any evaluation split; 0 accession overlap between splits; `test2` labels
 absent from train. Identical sequences carrying *different* EC labels may appear in
@@ -279,10 +262,10 @@ v1 limitations (all fixed in v2):
    `train`, because 1–2 cluster labels were split by sequence rather than by cluster
    (nucleotides: ~0.4%).
 
-See the repository / preprint for full details.
+See the GRIMM repo git history for full details.
 
 ## Citation
 
-> Hoarfrost et al. GRIMM: Genomic Representation Inference for Microbial Metabolism. (preprint)
+Preprint: https://arxiv.org/abs/2602.16504
 
 Code: https://github.com/Hoarfrost-Lab/grimm