You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

HDF5 nbit truncated chunk — silent wrong output

Overview

HDF5 nbit filter (H5Z_FILTER_NBIT, ID=5) accepts a chunked dataset whose compressed chunk has been truncated. H5Dread returns success (status=0) with no exception, but the materialized values differ from the expected values.

Files

File	Description
baseline_nbit.h5	Valid nbit-compressed HDF5 file; weights = [0..19]
crafted_nbit_truncated_chunk.h5	Same structure; nbit chunk truncated 80→20 bytes
reproduce.py	Reproduces silent wrong output via h5py
inspect_artifacts.py	Inspects file structure and source gap
runtime_results.json	C-harness runtime output
distinctness_matrix.json	Distinctness from F5 deflate and CVE-2024-32615
cve_2024_32615_diff.md	CVE diff analysis
SHA256SUMS.txt	File hashes

Reproduction

pip install h5py numpy
python reproduce.py

Expected: WRONG_ELEMENTS=15/20, EXCEPTION_RAISED=False

Non-Claims

No RCE, ACE, or memory corruption confirmed
No arbitrary read/write
No High/Critical/CVSS assertion
Not a duplicate of CVE-2024-32615 (distinct root, distinct function, distinct fix status)

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support