Add ablation chart, remove em dashes from model card

README.md

@@ -29,33 +29,35 @@ model-index:
             name: EM (EM-optimal gate)
           - type: precision
             value: 91.7
-            name: Answered precision @ 90%-precision gate (9% coverage)
+            name: Answered precision at 90%-precision gate (9% coverage)
 ---
 
 # Grounded Pointer QA
 
 An extractive question-answering model that **cannot hallucinate by
 construction**: its output layer can only point at spans inside retrieved
-passages of *your* documents — it has no vocabulary to generate from. A
-trained abstention head refuses when the loaded knowledge doesn't contain the
-answer, decoding is deterministic (argmax; same question + same documents =
-same answer forever), and knowledge is **hot-swappable**: point it at a new
-folder of text files and it answers from those, no retraining.
+passages of your documents. It has no vocabulary to generate from. A trained
+abstention head refuses when the loaded knowledge does not contain the answer,
+decoding is deterministic (argmax, so the same question over the same documents
+gives the same answer every time), and knowledge is **hot-swappable**: point it
+at a new folder of text files and it answers from those, with no retraining.
 
-Built on `roberta-base` (125M params) with pointer + abstention heads,
-finetuned on SQuAD v2 **against real TF-IDF retrieval** — the model never saw
-gold passages during training, only what the retriever actually returned, so
-it learned to abstain on retrieval misses too.
+Built on `roberta-base` (125M params) with pointer and abstention heads,
+finetuned on SQuAD v2 against real TF-IDF retrieval. The model never saw gold
+passages during training, only what the retriever actually returned, so it
+learned to abstain on retrieval misses too.
+
+![Exact match by build stage](ablation.png)
 
 ## Operating modes
 
 The checkpoint ships with a calibrated confidence gate
-(`P(answerable) × P(span)`), selected on a calibration split and verified on
-a held-out test split:
+(`P(answerable) x P(span)`), selected on a calibration split and verified on a
+disjoint held-out test split:
 
 | Mode | Gate | Coverage | Answered precision | EM |
 |---|---|---|---|---|
-| "Right or silent" (shipped default) | 0.965 | 9% | **91.7%** | 57.1 |
+| "Right or silent" (shipped default) | 0.965 | 9% | 91.7% | 57.1 |
 | EM-optimal | 0.295 | 45% | 72.3% | 74.6 |
 
 Pass a lower `gate` to `ask()` for more coverage at lower precision.
@@ -64,7 +66,7 @@ Pass a lower `gate` to `ask()` for more coverage at lower precision.
 
 ```python
 # files needed: proqa.pt, modeling_proqa.py (both in this repo)
-# pip install torch transformers scikit-learn numpy  (+ pypdf for PDFs)
+# pip install torch transformers scikit-learn numpy  (plus pypdf for PDFs)
 from modeling_proqa import GroundedQA
 
 qa = GroundedQA("proqa.pt")
@@ -82,16 +84,16 @@ qa.ask("what is the capital of France?")   # not in your docs
 
 - Single-span extraction only: no summarization, no aggregation across
   passages, no multi-turn conversation.
-- Trained on Wikipedia-style prose; **tables read poorly** — convert rows to
-  sentences ("ATA chapter 32 is Landing Gear.") for big accuracy gains.
-- TF-IDF retrieval is lexical: paraphrases sharing no words with your
+- Trained on Wikipedia-style prose, so **tables read poorly**. Convert rows to
+  sentences ("ATA chapter 32 is Landing Gear.") for large accuracy gains.
+- TF-IDF retrieval is lexical, so paraphrases sharing no words with your
   documents may cause (safe) abstentions.
-- Single training run, single seed; in-domain calibration.
+- Single training run, single seed, in-domain calibration.
 
-Full writeup, ablations (from-scratch 22.4 EM → 74.6 EM ladder), and negative
-results: see the project repository's `paper/`.
+Full writeup, ablations (the 22.4 to 74.6 EM ladder above), and negative
+results are in the project repository's `paper/`.
 
 ## Training
 
-One NVIDIA RTX 5060 Ti (16 GB): ~2.5 h finetune (batch 8 × 4 passages × 384
-tokens, bf16, lr 2e-5, 2 epochs) + calibration pass.
+One NVIDIA RTX 5060 Ti (16 GB): about 2.5 h finetune (batch 8 x 4 passages x
+384 tokens, bf16, lr 2e-5, 2 epochs) plus a calibration pass.