Identify all field types in CUAD dataset#25
Open
anaynasingh wants to merge 5 commits into
Open
Conversation
Extracted the 5 documents with the most annotated clauses from CUADv1.json for use in the loan extractor engine. https://claude.ai/code/session_01LUzUuGk96w6XUAcus9Gcrs
Organises rough session notes into structured action items across 13 areas: multi-span schema, confidence display, metrics, scoring systems, FP/FN analysis, field QA, dashboard improvements, seed data, and dataset expansion. https://claude.ai/code/session_01LUzUuGk96w6XUAcus9Gcrs
Covers root causes of poor scores (multi-span schema mismatch + narrow field descriptions), improved descriptions for 4 problem fields, and prioritised fix list with concrete examples from CUAD GT. https://claude.ai/code/session_01LUzUuGk96w6XUAcus9Gcrs
9 attack contracts + 1 clean baseline covering: direct instruction override, fake system prompt blocks, role hijack, system prompt exfiltration, confidence inflation, hidden HTML comment injection, fake assistant turn, tab-hidden injection, and schema format override. Includes automated test runner. https://claude.ai/code/session_01LUzUuGk96w6XUAcus9Gcrs
…nner Layer 2: schema type checks, value length guards, unknown field detection, and injection keyword scan across all field values — catches unknown attacks without needing to know the payload upfront. Layer 3: Claude-as-judge reads contract + extraction and flags manipulated output. Opt-in via --layer3 flag, requires --anthropic-key. https://claude.ai/code/session_01LUzUuGk96w6XUAcus9Gcrs
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Extracted the 5 documents with the most annotated clauses from CUADv1.json for use in the loan extractor engine.
https://claude.ai/code/session_01LUzUuGk96w6XUAcus9Gcrs