Upload _reasoning_rl_multiseed/eval_reasoning_t3_v7r128_postRL_alpha1_s2_par_20260506_221253/score.md with huggingface_hub
Browse files
_reasoning_rl_multiseed/eval_reasoning_t3_v7r128_postRL_alpha1_s2_par_20260506_221253/score.md
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Reasoning quality metrics · n=50
|
| 2 |
+
|
| 3 |
+
## Format / sanity
|
| 4 |
+
|
| 5 |
+
| metric | value |
|
| 6 |
+
|---|---:|
|
| 7 |
+
| reasoning_tags_rate | 0.6400 |
|
| 8 |
+
| dna_tags_rate | 0.0000 |
|
| 9 |
+
| cell_mention_rate | 0.7800 |
|
| 10 |
+
| mean_len_chars | 907.2000 |
|
| 11 |
+
| p10_len_chars | 562.9000 |
|
| 12 |
+
| p90_len_chars | 1381.2000 |
|
| 13 |
+
|
| 14 |
+
## TF-grounding (the headline reasoning metric)
|
| 15 |
+
|
| 16 |
+
| metric | value |
|
| 17 |
+
|---|---:|
|
| 18 |
+
| mean_tf_grounded_rate | 0.2397 |
|
| 19 |
+
| mean_n_cited | 14.3800 |
|
| 20 |
+
| mean_n_grounded | 6.4600 |
|
| 21 |
+
| mean_n_canonical_only | 0.4000 |
|
| 22 |
+
| mean_n_hallucinated | 7.5200 |
|
| 23 |
+
| mean_motif_consensus | 0.3865 |
|
| 24 |
+
| mean_prog_coverage | 0.1130 |
|
| 25 |
+
|
| 26 |
+
## Per-cell breakdown
|
| 27 |
+
|
| 28 |
+
| cell | n | tf_grounded_rate ↑ | n_cited | len_chars | cell_mention |
|
| 29 |
+
|---|---:|---:|---:|---:|---:|
|
| 30 |
+
| Ast | 5 | 0.199 | 12.0 | 789 | 1.000 |
|
| 31 |
+
| End | 3 | 0.333 | 5.3 | 1102 | 0.000 |
|
| 32 |
+
| Ex | 15 | 0.366 | 23.3 | 1032 | 0.800 |
|
| 33 |
+
| In | 4 | 0.056 | 8.2 | 998 | 1.000 |
|
| 34 |
+
| Mic | 13 | 0.173 | 11.4 | 828 | 0.769 |
|
| 35 |
+
| OPC | 2 | 0.167 | 7.5 | 972 | 0.500 |
|
| 36 |
+
| Oli | 8 | 0.211 | 12.2 | 743 | 0.875 |
|