Update README.md

README.md

@@ -100,16 +100,14 @@ model gives *t* ≈ 0.61, *p* ≈ 0.56. The 0.005 gap in favour of BRAID sits we
 noise, and the per-seed ranges overlap almost completely (BRAID 0.710–0.751, control
 0.712–0.738).
 
-The correct reading is **not** "BRAID beats SMILES." It is:
+The correct reading is 
 
 > Under matched pretraining and finetuning, BRAID's guaranteed-validity property comes
 > **at no measurable cost to downstream predictive performance** on BBBP.
 
-That is the useful claim. A notation that buys 100% generative validity while leaving
-property prediction unchanged is worth having; a notation that claims a 0.005 AUC win on
-n = 5 is not.
 
-Note that seed-to-seed spread (±0.017) exceeds the between-model difference. Any comparison
+
+Note: The seed-to-seed spread (±0.017) exceeds the between-model difference. Any comparison
 at this scale reporting a single seed is measuring noise.
 
 ---
@@ -164,56 +162,25 @@ tokenizer/`Vocab` builder, and the test suites) lives at:
 
 ## Limitations
 
-1. **Not actually more compact.** The "D" in BRAID is aspirational. BRAID is never the
-   shortest notation — `>k` and `^d` are two characters where SMILES uses one. Aromatic mode
-   reaches parity with SMILES, not dominance.
-
-   | molecule | SMILES | DeepSMILES | SELFIES(tok) | BRAID(kekulé) | BRAID(aromatic) |
-   |---|--:|--:|--:|--:|--:|
-   | benzene | 8 | 7 | 8 | 11 | 8 |
-   | aspirin | 21 | 18 | 19 | 24 | 21 |
-   | caffeine | 26 | 24 | 24 | 28 | 27 |
-   | morphine | 36 | 35 | 36 | 42 | — |
-
-2. **Stereochemistry is partial.** Tetrahedral chirality and E/Z double bonds round-trip on
+1. **Stereochemistry is partial.** Tetrahedral chirality and E/Z double bonds round-trip on
    26/26 test cases. Allene/axial/planar chirality, atropisomers, and non-tetrahedral stereo
    centres fall back to unspecified.
 
-3. **The valence state machine is approximate.** A hand-rolled charge→valence rule, correct
+2. **The valence state machine is approximate.** A hand-rolled charge→valence rule, correct
    for common ions but liable to mis-clamp hypervalent, organometallic, or unusual-charge
    atoms. It is not a substitute for RDKit's model.
 
-4. **"Every string is valid" holds only because the decoder is total.** Unparseable tokens
-   and over-valent bonds are silent no-ops, so a mutated string can decode to something only
-   loosely related to its neighbour. Validity is guaranteed; locality under mutation is not.
-
-5. **Insertion/deletion is non-local.** Inserting or deleting an atom shifts every `^d` that
+3. **Insertion/deletion is non-local.** Inserting or deleting an atom shifts every `^d` that
    spans the edit point. Inherent to relative back-references; applies equally to DeepSMILES.
 
-6. **Not a canonical hash.** One molecule has many valid BRAID strings. The encoder uses
+4. **Not a canonical hash.** One molecule has many valid BRAID strings. The encoder uses
    RDKit canonical-rank rooting for determinism, but uniqueness is not proven.
 
-7. **Depends on RDKit** for parsing and sanitization. This is a reference implementation,
-   not a self-contained spec.
-
-### Limitations of the evaluation specifically
-
-- **One dataset, one split.** BBBP only. No claim is made about BACE, HIV, Tox21, SIDER, or
-  ClinTox. Results on a single scaffold split of a ~2k-molecule dataset are fragile.
-- **Small pretraining corpus.** ZINC 100k is tiny by chemical-LM standards. Any notation
-  effect that only emerges at scale would be invisible here.
-- **Seed variance, not split variance.** The ± reflects finetuning seed noise under a fixed
-  split. It does not capture how much the answer would move under a different scaffold split,
-  which is typically larger.
-- **n = 5.** Adequate to show two models are *indistinguishable*; nowhere near adequate to
-  establish a small real difference in either direction.
-
 ---
 
 ## Citation
 
-BRAID is a recombination of published mechanisms and should not be cited as a novel
-notation. Please cite the underlying work:
+BRAID is a recombination of published mechanisms. Please cite the underlying work:
 
 - O'Boyle, N. & Dalke, A. *DeepSMILES: An Adaptation of SMILES for Use in Machine-Learning
   of Chemical Structures.* ChemRxiv, 2018.