wadhma commited on
Commit
21d9c9c
·
verified ·
1 Parent(s): 5e8232a

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +14 -0
README.md CHANGED
@@ -1,5 +1,9 @@
1
  ---
2
  configs:
 
 
 
 
3
  - config_name: rat_positions
4
  data_files:
5
  - split: train
@@ -98,6 +102,7 @@ git sha(s): 3df492e12210930d6418b45f2ec1cd9a6f05ee42, 3df492e12210930d6418b45f2e
98
 
99
  Rollouts: 133400 rows written (133400 successful after retries, 0 still failed). finish_reason counts: {'stop': 133400}.
100
  Positions: 1334 sentences over 10 instances, 0 refused.
 
101
 
102
  Method: Thought Anchors resampling importance. For sentence i, the set is `n_rollouts` continuations from the trace prefix before s_i; importance is KL(P_i || P_{i+1}) over bucketed answers.
103
  Model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), thinking on.
@@ -152,4 +157,13 @@ git sha(s): 3df492e12210930d6418b45f2ec1cd9a6f05ee42. Full manifest: manifests/r
152
  | prefix_chars / prompt_tokens / max_tokens | prefix length (chars, tokens) and the cap used |
153
  | status / reason | ok, or refused (not enough room in the context window) with the reason |
154
 
 
 
 
 
 
 
 
 
 
155
  <!-- /config:rat -->
 
1
  ---
2
  configs:
3
+ - config_name: rat_importance
4
+ data_files:
5
+ - split: train
6
+ path: data/rat_importance.jsonl
7
  - config_name: rat_positions
8
  data_files:
9
  - split: train
 
102
 
103
  Rollouts: 133400 rows written (133400 successful after retries, 0 still failed). finish_reason counts: {'stop': 133400}.
104
  Positions: 1334 sentences over 10 instances, 0 refused.
105
+ Importance: 1324 rows, 1324 scored, alpha=0.1, mean 0.0858.
106
 
107
  Method: Thought Anchors resampling importance. For sentence i, the set is `n_rollouts` continuations from the trace prefix before s_i; importance is KL(P_i || P_{i+1}) over bucketed answers.
108
  Model: Qwen/Qwen3-8B (revision b968826d9c46dd6066d109eabc6255188de91218), thinking on.
 
157
  | prefix_chars / prompt_tokens / max_tokens | prefix length (chars, tokens) and the cap used |
158
  | status / reason | ok, or refused (not enough room in the context window) with the reason |
159
 
160
+ ### rat_importance columns
161
+ | Column | Description |
162
+ |---|---|
163
+ | counts | bucketed answer counts of the set at this sentence (<no_answer> = no extractable answer) |
164
+ | n_ok / n_failed / n_no_answer | completions used / failed requests / no extractable answer |
165
+ | top_answer | most common bucket |
166
+ | importance | KL(P_i || P_{i+1}); null for the last sentence or a missing set |
167
+ | alpha | smoothing used |
168
+
169
  <!-- /config:rat -->