synthdatasets / splittingVersions /RESULT_all_v4.md
tkhersoft's picture
splitting versions (unbundle/out + RESULT_all_v*.md)
7e06839 verified
|
Raw
History Blame Contribute Delete
1.25 kB
# Unbundle v4 (auth handshake + humanized): C3 Results
Baseline (n100, single-turn): **514/795 = 64.7%** caught
Unbundled dataset: 1299 rows judged (0 errors)
Overall caught: **554/1299 = 42.6%**
## Row composition
- turn1 (preserved trie divergence): 282
- tail turns (one action each): 504
- untouched passthrough: 513
## T1 slots vs baseline (same 283 trie positions)
- baseline caught here: 223/282
- unbundled caught here: 54/282 = 19.1%
- **fixed (caught -> fooled): 169**
- regressed (fooled -> caught): 0
- net catch reduction on T1: 169
## Tail turns (the re-rooted follow-ups)
- tail caught: 207/504 = 41.1%
- by original tail length:
tail_len=1: 38/133 = 29%
tail_len=2: 65/154 = 42%
tail_len=3: 100/213 = 47%
tail_len=4: 4/4 = 100%
- by turn position:
turn2: 92/282 = 33%
turn3: 69/149 = 46%
turn4: 45/72 = 62%
turn5: 1/1 = 100%
## Passthrough control (should track baseline)
- caught: 293/513 = 57.1%
- same rows in baseline: 291/513 = 56.7%
## Per-scenario view (283 split confusers)
- caught in >=1 of its turns: 160/282 = 56.7%
- caught at the T1 (trie) slot: 54/282 = 19.1%
- baseline caught these same confusers: 223/282 = 79.1%