Non-autoregressive System 1 decision engine. Typed choice, score and yes/no decisions over any text in a single forward pass, in 100+ languages, with a router that picks the right checkpoint per request.
Python calibrationclassificationdecision-modelhuggingfacejevmodernbertmultilingualnlppythonpytorchroutingtyped-decisionszero-shot
Merge pull request #476 from humairaambreen/docs/confidence-option-count-caveat
Nandakishor6af329986bca
diff --cc README.mdindex 1b45b288,93956c4c..a48d34cd--- a/README.md+++ b/README.md@@@ -544,9 -451,9 +544,9 @@@ the router auto-selects by script/languThree things differ from Jev when you port a client:-* **Options per question.** A question's options share the checkpoint's option budget, `head_max_len` (192 tokens on `laya`, 256 on the other two), not Jev's cap of 255 options. Once they overflow it, around 20 options with a short description each, every option is trimmed to fit, so long or similar labels can reach the model reading the same ([Where Jev leads](#where-jev-leads)). Once they no longer fit the window at all, the request is rejected with 422. With short labels such as `Queue 042: Tickets routed to queue 42` that happens above 126 options on `laya` and 254 on the other two; the exact point moves with the length of the instructions and labels. For more candidates, narrow them first with `predict_shortlist` ([Honest limits](#honest-limits)).+* **Options per question.** A question's options share the checkpoint's option budget, `head_max_len` (192 tokens on `laya`, 256 on the other two), not Jev's cap of 255 options. In addition, the HTTP server (`laya.serve`) enforces an amplification guard of at most 100 choice options per question (`MAX_CHOICE_OPTIONS = 100`, rejected with 413 before inference). Once options overflow the token budget, around 20 options with a short description each, every option is trimmed to fit, so long or similar labels can reach the model reading the same ([Where Jev leads](#where-jev-leads)). Once they no longer fit the window at all, the library rejects the request with 422. For more candidates, narrow them first with `predict_shortlist` ([Honest limits](#honest-limits)).* **Score levels.** Every level needs a description. A `null` level is rejected with 422 rather than scored and echoed back in `legend`.- * **`confidence`** on `choice` and `score` answers is 1 minus normalised entropy, a measure of how concentrated the distribution is, not Jev's `(n·p_max − 1)/(n − 1)`. A threshold carried over from Jev does not transfer. For one calibrated number on every question type, gate on `answer_confidence`, the probability of the reported answer.+ * **`confidence`** on `choice` and `score` answers is 1 minus normalised entropy, a measure of how concentrated the distribution is, not Jev's `(n·p_max − 1)/(n − 1)`. A threshold carried over from Jev does not transfer. For one calibrated number on every question type, gate on `answer_confidence`, the probability of the reported answer. Both values shift with the number of options, so fit and validate any cutoff on held-out data at the option counts your workload uses ([#394](https://github.com/NandhaKishorM/laya/issues/394)).### Nix / NixOS