SoulInPsyAbstract commited on
Commit
2e90020
·
verified ·
1 Parent(s): fc8a3ed

Add v2 control-run caveat (15/20 refusals, 5/20 fabrications, not 20/0)

Browse files
Files changed (1) hide show
  1. README.md +6 -0
README.md CHANGED
@@ -19,6 +19,12 @@ metrics:
19
 
20
  **20/20 refusals. 0/20 fabrications.**
21
 
 
 
 
 
 
 
22
  Qwen2.5-7B-Instruct fine-tuned on the Protocol 0 Binary dataset.
23
 
24
  Same data, same 3 epochs, same binary format. Result: perfect abstention without a single fabricated number.
 
19
 
20
  **20/20 refusals. 0/20 fabrications.**
21
 
22
+ > **Update (2026-08-03):** the 20/20 / 0/20 numbers above used a scorer that only checked whether
23
+ > the response started with "TRUE"/"FALSE", and could not detect a fabricated number stated
24
+ > anywhere else in the response — an artifact, not a comparable measurement. A v2 control run
25
+ > (30 tokens, one money-regex scorer applied identically to base and fine-tuned models) gives
26
+ > **15/20 refusals, 5/20 fabrications** for this model. Raw results: [binary_sft_k20_v2.json](https://huggingface.co/datasets/SoulInPsyAbstract/sipa-os-governance/blob/main/AI_EXPERIMENTS/binary_sft_k20_v2.json).
27
+
28
  Qwen2.5-7B-Instruct fine-tuned on the Protocol 0 Binary dataset.
29
 
30
  Same data, same 3 epochs, same binary format. Result: perfect abstention without a single fabricated number.