mathiasandClaude Sonnet 5.5 5e54c9beff feat(reconcile): LLM adjudicator with the measured prompt, plus an order-checked wrapper ...

Sends the exact wording that scored 92% (gemma4-26b) and 97% (gemma4-31b); maps
1/2 back to base/alt under both orders; an unreadable reply is unsure, a transport
error is an error. OrderChecked asks twice with the order swapped and keeps only
agreement. Mutation-checked: answer mapping, unparseable-as-guess, the word
boundary and the order check each turn a test red.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

2026-10-01 23:18:11 +02:00

..

adjudicate_test.go

test(reconcile): a number or word that merely starts with 1 or 2 is not an answer

2026-10-01 23:18:11 +02:00

adjudicate.go

feat(reconcile): LLM adjudicator with the measured prompt, plus an order-checked wrapper

2026-10-01 23:18:11 +02:00

diff_test.go

test(reconcile): word-level diff is minimal, tiles both inputs and is fast (red)

2026-10-01 23:15:44 +02:00

diff.go

feat(reconcile): minimal word-level diff (Myers) with a bounded edit distance

2026-10-01 23:16:31 +02:00

disputes_test.go

test(reconcile): finding disputes, resolving the trivial ones, and their context (red)

2026-10-01 23:16:55 +02:00

disputes.go

feat(reconcile): find disputes, resolve trivial ones without a model, give context

2026-10-01 23:17:21 +02:00

tokens_test.go

test(reconcile): tokenising for word-level comparison (red)

2026-10-01 23:15:44 +02:00

tokens.go

feat(reconcile): tokenise for word-level comparison

2026-10-01 23:16:31 +02:00