CVE-2026-44223

vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters

Published May 12, 2026 CVSS 3.1 · 6.5 Medium EPSS < 1% KEV · Not listed
Description
vLLM is an inference and serving engine for large language models (LLMs). From 0.18.0 to before 0.20.0, the extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. This vulnerability is fixed in 0.20.0.

Analysis

AI-generated

Technical impact

A bug in the extract_hidden_states speculative decoding proposer causes the function to return a tensor with an incorrect shape after the first decode step, which triggers a RuntimeError that terminates the EngineCore process. This flaw results from a combination of size handling errors and type conversion mistakes, consistent with CWE‑131 (incorrect size handling) and CWE‑704 (incorrect type conversion). The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., repetition_penalty = 1.1) is sufficient to crash the server. This vulnerability is fixed in version 0.20.0. The flaw does not expose any arbitrary code execution or data exfiltration capabilities; it only results in service unavailability for any affected session.

Affected Systems

The vLLM inference engine, maintained by the vllm-project under the product name vllm, is affected. Versions of the software released prior to 0.20.0 are vulnerable; the 0.20.0 release and later versions the fix.

Risk and Exploitability

The CVSS score of 6.5 indicates a moderate impact, and the EPSS score of 0.367% (reported as <1%) indicates a very low likelihood of exploitation. The vulnerability is not listed in the CISA KEV catalog. Attackers can trigger the crash by sending a single request with a penalty parameter, which can be performed remotely against any accessible vLLM instance. The only requirement is the ability to submit a request payload that includes one of the penalty parameters; no additional privileged access is needed.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Update vLLM to the 0.20.0 release or a later version that contains the patch for the tensor shape bug.
  • Temporarily disable or filter out the penalty parameters (repetition_penalty, frequency_penalty, presence_penalty) in request payloads until the patch is applied.
  • Monitor the application logs for RuntimeError messages or sudden process terminations to detect exploitation attempts early.

Generated by OpenCVE AI on June 23, 2026 at 00:21 UTC.

Tracking

Sign in to view the affected projects.

Advisories

Source ID Title
Github GHSA Github GHSA GHSA-83vm-p52w-f9pw vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters
History
Type Values Removed Values Added
Description vLLM is an inference and serving engine for large language models (LLMs). From to before 0.20.0, the extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. This vulnerability is fixed in 0.20.0. vLLM is an inference and serving engine for large language models (LLMs). From 0.18.0 to before 0.20.0, the extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. This vulnerability is fixed in 0.20.0.
Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'no', 'Exploitation': 'poc', 'Technical Impact': 'partial'}, 'version': '2.0.3'}

Type Values Removed Values Added
First Time appeared Vllm
Vllm vllm
CPEs cpe:2.3:a:vllm:vllm:*:*:*:*:*:*:*:*
Vendors & Products Vllm
Vllm vllm
Type Values Removed Values Added
First Time appeared Vllm-project
Vllm-project vllm
Vendors & Products Vllm-project
Vllm-project vllm
Type Values Removed Values Added
Description vLLM is an inference and serving engine for large language models (LLMs). From to before 0.20.0, the extract_hidden_states speculative decoding proposer in vLLM returns a tensor with an incorrect shape after the first decode step, causing a RuntimeError that crashes the EngineCore process. The crash is triggered when any request in the batch uses sampling penalty parameters (repetition_penalty, frequency_penalty, or presence_penalty). A single request with a penalty parameter (e.g., "repetition_penalty": 1.1) is sufficient to crash the server. This vulnerability is fixed in 0.20.0.
Title vLLM: extract_hidden_states speculative decoding crashes server on any request with penalty parameters
Weaknesses CWE-131
CWE-704
References
Metrics cvssV3_1

{'score': 6.5, 'vector': 'CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:N/I:N/A:H'}

Weaknesses
  • CWE-131

    Incorrect Calculation of Buffer Size

  • CWE-704

    Incorrect Type Conversion or Cast

Sources

Status
PUBLISHED
Assigner
GitHub_M
Published
Updated
Jun 22, 2026, 9:49 PM
Reserved
May 5, 2026, 3:42 PM
Status
Modified
Published
May 12, 2026, 8:16 PM
Modified
Nov 11, 2026, 8:10 PM