bhaskargurram-ai/verifydoc - 20 Help Wanted Issues

Open Issues Need Help

Add >=3 local/open extractors: PaddleOCR-VL, dots.ocr (vLLM), and open VLMs 10 days ago

help wanted adapters

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Learned-fusion feature importance / interpretability 10 days ago

good first issue eval

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Guarantee-respecting genuine-VLM conformal (nominal alpha, doc-clustered, multiplicity-corrected) 10 days ago

help wanted calibration

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Ablation: does provenance beat non-provenance conditioning covariates? 10 days ago

help wanted calibration

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Trust-gated agentic layer: accuracy-at-fixed-compute evaluation 10 days ago

enhancement help wanted

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Alternative ambiguity-penalty modes (softmax / entropy) 10 days ago

good first issue calibration

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Document-clustered CIs + across-split dispersion on every headline table 10 days ago

help wanted eval

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Grow VerifyDocBench: DocILE, SROIE, full multilingual XFUND, invoices 10 days ago

help wanted benchmark

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Formalize the "when does provenance-conditioning help" characterization 10 days ago

enhancement help wanted

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Human-labeled gold slice + inter-annotator agreement for free-text fields 10 days ago

help wanted benchmark

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

CLI test coverage (tests/test_cli.py) 13 days ago

good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

MCP registry + awesome-mcp-servers submission 13 days ago

good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Batch/folder processing CLI: verifydoc batch <dir> 13 days ago

good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Array-leaf alignment (greedy matching) in JSON scoring 13 days ago

good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

parse_number: strip alphabetic currency codes (RM, Rp, USD) 13 days ago

bug good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Adaptive-k budget control for consensus confidence 13 days ago

enhancement good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

DocILE + XFUND dataset loaders 13 days ago

enhancement good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

Loaders + license audit: FUNSD, SROIE, DocILE, XFUND 14 days ago

enhancement good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

VerifyDocBench v1: CORD slice with gold source boxes 14 days ago

enhancement good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai

README demo GIF: Streamlit click-through on a hallucinated field 14 days ago

documentation good first issue

bhaskargurram-ai/verifydoc

Trust layer for document→JSON extraction: calibrated per-field confidence, source grounding, and abstention on top of any extractor (+ VerifyDocBench eval harness)

Python

#benchmark#calibration#conformal-prediction#document-ai#document-extraction#hallucination-detection#llm#ocr#selective-prediction#trustworthy-ai