Commit History
21 days ago
-
Model card: wording
Verified signature93ee47b
29 days ago
-
Card: drop the note on where D1A is served
Verified signature44db470 -
Card: drop the notes on where D1A is served
Verified signaturec85137f
30 days ago
-
Card: capabilities table, where each model runs since 10-14; E2B-it's suites measured
Verified signature11e39a4
-
Card: capabilities table: chat and embeddings rows (d1a#276)
Verified signature8465fcd -
Card: a capabilities table of every D1A model line (jonpol01/d1a#273)
Verified signaturec808e3b
-
v0.6.1: routing temperature (use_case_temperatures.routing), same weights as v0.6 (jonpol01/d1a#264)
Verified signature87f2b75
-
Card: the chosen D1A avatar
Verified signature6c0ebe2 -
Card: add the D1A avatar
Verified signatured143fb9
-
D1A-E4B v0.6
Verified signature3ca01b8
-
D1A-E4B v0.5: Kev's skills back on top of v0.4 (hard 55->71, devtools 64->70); PR blast/severity trade-off in the card
Verified signaturee7db32f
-
v0.4: pull-request labeling round 2
Verified signaturedbd2df8
-
v0.3: pull-request labeling (English and Japanese)
Verified signaturec9bf9f6 -
D1A-E4B v0.2 (hybrid): + Kev recipe stages + Japanese JGLUE + agent-factory/routing; calibrated T=1.95
Verified signaturee8064af
-
Calibration temperature 1.46 (fitted on the study's calibration split; the published head.pt had been left at 1.0)
Verified signature2030f21 -
D1A-E4B v0.1: two epochs at lr 5e-5 (dev acc 0.853 trained / 0.680 new sources)
Verified signature98847a9
-
Reserve D1A: placeholder model card
Verified signature3a2fcec -
initial commit
Verified signature69e02d2