card wording

README.md

@@ -74,7 +74,7 @@ Kev-4B is a **decision model**: one document (the *state*) and a set of typed qu
 
 Per-source out-of-domain accuracy (Kev-4B / Jev): QNLI 0.91 / 0.93, SciQ 0.97 / 0.99, TweetEval-offensive 0.74 / 0.81, PAWS 0.74 / 0.79, MMLU 0.70 / 0.90, Emotion 0.56 / 0.59, deadline (3-level date arithmetic) 0.60 / 0.93 — **0.85 with the `date_facts` preprocessor** (below), (A or B) and C 0.91 / 0.91, (A and B) or not C 0.88 / 0.97, if A then not B else C 1.00 / 0.78.
 
-**Calibration is built in.** `head.pt` carries a temperature (T = 2.14) fitted on this checkpoint's in-distribution development rows by minimising negative log-likelihood ([`scripts/calibrate_checkpoint.py`](https://github.com/jaredpalmer/kev/blob/main/scripts/calibrate_checkpoint.py)); the pointer head divides its logits by it at inference. Every loader — `kev.serve`, `kev.benchmark`, the Space, anyone's harness — gets the calibrated probabilities by default. It never changes an answer: the argmax and the confidence ranking are identical, so accuracy and coverage are the same in both columns. `KEV_TEMPERATURE=1.0` restores the raw logits; the raw column is what the training produced. Per-(type, option-count) temperatures were tested and are worse out of domain. The fit uses no out-of-domain or test data.
+**Calibration is built in.** `head.pt` carries a temperature (T = 2.14) fitted on this checkpoint's in-distribution development rows by minimising negative log-likelihood ([`scripts/calibrate_checkpoint.py`](https://github.com/jaredpalmer/kev/blob/main/scripts/calibrate_checkpoint.py)); the pointer head divides its logits by it at inference. Every loader — `kev.serve`, `kev.benchmark`, the Space, anyone's harness — gets the calibrated probabilities by default. It never changes an answer: the argmax is identical, so accuracy is the same in both columns; confidences are re-ordered only slightly across questions with different option counts, which is why coverage moves by a point or two. `KEV_TEMPERATURE=1.0` restores the raw logits; the raw column is what the training produced. Per-(type, option-count) temperatures were tested and are worse out of domain. The fit uses no out-of-domain or test data.
 
 **`date_facts` preprocessor.** Kev, like every Kev before it, cannot subtract dates reliably (the untrained base can; LoRA training erodes it). It can use a stated day count. `KEV_DATE_FACTS=1` appends one sentence per pair of absolute dates found in the state ("June 26, 2026 is 8 days before July 4, 2026"); this checkpoint was trained on such renderings, so with it `deadline` goes from 0.60 to 0.85 and overall out-of-domain accuracy from 0.797 to 0.820. It is preprocessing, reported separately, never folded into the model's own numbers.