Upload README.md with huggingface_hub

README.md

@@ -31,8 +31,6 @@ expA/{topic}/{benchmark}/               # Exp A: influence-guided, per-topic for
                transportation, travel_and_tourism
   benchmarks: gsm8k | mmlu_social_science | mmlu_stem | socialiqa
 
-expB/{group}/                           # Exp B: influence-guided, multi-topic groups
-  groups: gsm8k_A/B/C | mmlu_stem_A/B/C | socialiqa_B/C
 ```
 
 > **Note:** Each checkpoint is the *last healthy* checkpoint — the last regular save
@@ -99,7 +97,6 @@ model = model.merge_and_unload()
 | Exp 1 / Exp 3 | Random sample from DOLMA-3 6T | None |
 | Exp C | Top-2000 docs by TracStar influence score | None |
 | Exp A | Top-2000 docs by TracStar influence score | Per topic (24 topics) |
-| Exp B | Top-2000 docs by TracStar influence score | Multi-topic groups |
 
 Influence scores are from **TracStar** (training data attribution) computed against
 four evaluation benchmarks: GSM8K, MMLU Social Science, MMLU STEM, SocialIQA.