Richie Cahill 251da6c14a added bill_token_compression.py ...

tested on sample size of 100 bills matching the distribution of our data
Compression saves ~11.5% on prompt tokens; completion/reasoning are roughly equal across the two sets.
prompt	completion	reasoning	total
compressed	349,460	157,110	112,128	506,570
uncompressed	394,948	154,710	110,080	549,658
delta	−45,488	+2,400	+2,048	−43,088

2026-04-09 18:41:13 -04:00

..

__init__.py

creating prompt_bench downloader

2026-04-07 19:15:43 -04:00

bill_token_compression.py

added bill_token_compression.py

2026-04-09 18:41:13 -04:00

compresion_test.py

added bill_token_compression.py

2026-04-09 18:41:13 -04:00

container.py

created main prompt bench

2026-04-08 09:08:25 -04:00

downloader.py

creating prompt_bench downloader

2026-04-07 19:15:43 -04:00

main.py

created main prompt bench

2026-04-08 09:08:25 -04:00

models.py

created main prompt bench

2026-04-08 09:08:25 -04:00

vllm_client.py

created main prompt bench

2026-04-08 09:08:25 -04:00