Deserialization in Hiyouga Llama-Factory

CVE-2025-46567

LLama Factory enables fine-tuning of large language models. Prior to version 1.0.0, a critical vulnerability exists in the `llamafy_baichuan2.py` script of the LLaMA-Factory project. The script performs insecure deserialization using `torch.load()` on user-supplied `.bin` files from an input directory. An attacker can exploit this behavior by crafting a malicious `.bin` file that executes arbitrary commands during deserialization. This issue has been patched in version 1.0.0.

Vulnerability class: Insecure Deserialization

EPSS: 0.002 (16.0th percentile) — read the EPSS interpretation.

CVSS v3 metric

CVSS v3 base score 6.1 (Medium). Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:R/S:U/C:H/I:L/A:L.

Affected products

Weakness classification (CWE)

Public proof-of-concept exploits

References

Frequently asked questions

What is CVE-2025-46567?
CVE-2025-46567 is a medium-severity vulnerability in Hiyouga Llama-Factory, classified under Deserialization of Untrusted Data. CVSS score: 6.1/10. Published 2025-05-01.
How severe is CVE-2025-46567?
Medium severity. CVSS v3 base score is 6.1 out of 10.
Is CVE-2025-46567 known to be exploited?
1 public proof-of-concept repositories are indexed. Not currently listed in the CISA KEV catalog.