Upload README.md with huggingface_hub

README.md

@@ -27,7 +27,7 @@ met.
 
 The training corpus contains 500,000 deduplicated structural examples grounded
 in 5,256 independently generated and audited tasks from 141 real-site domains.
-Domains—not rows—define the split: 110 train, 8 validation, and 23 untouched
+Domains—not rows—define the split: 110 train, 8 validation, and 23 held-out
 test domains.
 The corpus is primarily BetterWright accessibility trees, with a smaller share
 of alternate accessibility serializations for robustness.
@@ -87,11 +87,25 @@ fallback. A raw relevance score alone is not a safe deletion decision.
 
 ## Evaluation policy
 
-Fallback thresholds are selected only on validation domains. The public test
-report measures exact target/evidence-node recall, complete-task node recall,
-fallback rate, and tokenizer-measured savings on the untouched test domains.
-The report and production threshold file are published after the checkpoint's
-post-training quality gate; claims are not inferred from training loss.
+Fallback thresholds are selected on validation domains. The production policy
+uses eight ranked 1,800-character structural regions, preserves referenced
+nodes, bounded child context, ancestry, and deterministic must-retain states,
+and returns the exact original tree on low confidence, errors, insufficient
+benefit, or an over-limit result.
+
+| Domain split | Tasks | Target/evidence recall | Perfect-task recall | Pruning coverage | Mean token savings when active | Token-weighted savings |
+| --- | ---: | ---: | ---: | ---: | ---: | ---: |
+| Validation (8 domains) | 288 | 100.000% | 100.000% | 12.50% | 33.38% | 3.42% |
+| Held-out test (23 domains) | 877 | 99.584% | 99.544% | 22.46% | 44.09% | 9.69% |
+| Post-training reserve (17 additional domains) | 161 | 100.000% | 100.000% | 33.54% | 16.78% | 2.42% |
+
+The exact reports and frozen policy are distributed as `eval.json`,
+`reserve_eval.json`, and `relevance_config.json`. The 99.5% target is evaluated
+on whole tasks and required accessibility refs, not inferred from training
+loss. A nine-site live navigational replay retained and clicked 6/9 current
+targets; two large pages exceeded the runtime ceiling after fallback and one
+site did not expose the requested label. This limitation is intentionally not
+hidden by substituting easier sites.
 
 ## Safety and limitations
 
@@ -99,3 +113,7 @@ This is a relevance model, not an autonomous browser agent. It must not be the
 only path for security decisions. Snapshot text is untrusted data. Consumers
 should keep deterministic must-retain rules and fall back to the full tree on
 model errors, timeouts, low confidence, or insufficient retained context.
+Savings are substantial when pruning activates, but coverage is deliberately
+conservative and very large pages can still fall back above the caller's size
+ceiling. This checkpoint is therefore an alpha integration, not a claim that
+every page becomes smaller.