Update README.md

README.md

@@ -55,15 +55,16 @@ All evaluations are zero-shot and use the [public Carbon evaluation pipeline](ht
 
 ### Downstream tasks
 
-| Category | Metric (%) | Carbon 8B | Carbon 3B | Evo2 7B (1M) |
+| Category | Metric (%) | Carbon 3B | Carbon 8B | Δ |
 |---|---|---|---|---|
-| Generative | SR eukaryote | **64.03** | <u>61.50</u> | 59.83 |
-| Variant effect prediction | BRCA2 AUROC | **85.60** | <u>84.64</u> | 83.52 |
-| | TraitGym Mendelian AUPRC by-chrom | <u>36.81</u> | 34.24 | **38.36** |
-| | ClinVar coding AUROC, 48 kb | <u>93.43</u> | 93.30 | **93.70** |
-| | ClinVar non-coding AUROC, 48 kb | **91.98** | <u>91.56</u> | 90.03 |
-| Perturbation | TATA v2 | <u>65.62</u> | **65.94** | 63.72 |
-| | SYN v2 | **92.18** | 82.78 | <u>84.92</u> |
+| Generative | Sequence Recovery eukaryote | 61.54 | **64.05** | +2.51 |
+| Variant effect prediction | BRCA2 | 84.63 | **85.72** | +1.09 |
+| | TraitGym Mendelian | 33.65 | **36.43** | +2.78 |
+| | ClinVar coding (24 kb) | 92.89 | **93.11** | +0.22 |
+| | ClinVar non-coding (24 kb) | 91.14 | **91.63** | +0.49 |
+| Perturbation | Nucleotide triplet-expansion | 85.20 | **89.05** | +3.85 |
+| | Synonymous codon | 88.89 | **91.46** | +2.57 |
+| Long-context retrieval | Genome-NIAH @ 393 kbp | 79.00 | **86.00** | +7.00 |
 
 ### Genome-NIAH (long-context retrieval)
 
@@ -82,6 +83,8 @@ Carbon-8B retrieves reliably up to its 32 k native boundary; **YaRN 4×** recove
 
 Generative modelling, variant-effect prediction, motif-perturbation analysis, and long-context retrieval on DNA sequences. For faster inference at shorter contexts, use **Carbon-3B**.
 
+⚠️ **Genetic data is highly sensitive.** Depending on how this model is used (local download, inference API/endpoints, third-party inference providers, Spaces demos or others), input and output data may be processed or handled differently by different providers or space owners. Please make sure you understand and agree with how your data is handled before using the model.
+
 ## License
 
 Apache 2.0.