Event date · · Ant Group

inclusionAI open-sourced Qwen3.5-4B-singprobe, a streaming guardrail probe for Qwen3.5-4B

Ant Group 蚂蚁集团Chinese AIOpen weights
FACT STATEMENT

inclusionAI released Qwen3.5-4B-singprobe, an Apache-2.0 licensed streaming guardrail probe built on Qwen/Qwen3.5-4B, on Hugging Face. It adds less than 0.5% decode-time overhead and scores query intent, response unsafety, and hallucination risk at every token.

China context

Original name
inclusionAI
Outside China
Open weights · huggingface.co
Claims
Company-reported; not yet independently evaluated
For builders
Developers outside China can download the Apache-2.0 weights from Hugging Face and integrate the probe via SGLang or vLLM branches.
For investors
The release shows Ant Group's inclusionAI is actively contributing open-weight safety tooling, which may influence enterprise adoption of Qwen models.
What happened

inclusionAI released Qwen3.5-4B-singprobe on Hugging Face under Apache-2.0. The model is an intrinsic streaming guardrail that reuses the base model's hidden states to score query intent, response unsafety, and hallucination risk at every token, adding less than 0.5% decode-time overhead. It is supported through SGLang and vLLM integration branches.

Technical significance

The probe taps layers [9, 19, 30] of Qwen3.5-4B and outputs 8 intents plus unsafe and hallucination scores. Reported F1 is 0.8650 for query intent classification, 0.8646 for response safety, R-AUC/T-AUC of 0.9870/0.9286 for streaming safety, and AUC 0.7855 for hallucination detection. Benign-response false-positive rate is 0.04% average across 5 datasets.

Industry impact

Developers deploying Qwen3.5-4B can add streaming safety and hallucination checks without running a separate safety model, reducing inference cost and latency. Competitors offering separate guard models face pressure to match this integrated approach.

Decision value

For enterprises using Qwen3.5-4B, this probe offers a low-overhead way to monitor and mitigate unsafe or hallucinated outputs in real time, potentially reducing compliance and reputational risk.

What to watch

Next signals to check: adoption of the SGLang/vLLM integration branches, independent benchmarks against the reported metrics, and whether inclusionAI extends SingProbe to other base models.

Latest in Chinese AI

  1. DeepSeekReportedly: OpenDesign team's OpenMotion lets DeepSeek Harness create motion graphics
  2. XiaomiReportedly: Xiaomi adds MiMo UltraSpeed to Token Plan Pro and Max for a limited time
  3. MiniMaxMiniMax publishes cryptanalysis findings from its internal AI model
  4. MiniMaxMiniMax publishes Cyber repository on GitHub
  5. TencentReportedly: Tencent opens Hy4 weights, completing the Hunyuan LLM open-weight rollout

All China AI Events

AIGC Newsletter

China AI, with sources and context.

Analysis of Chinese AI models, companies and policy, and what you can use outside China.