Skip to content
OpenTrain AIFor AI Companies

HFEPX · Eval paper review

Harder to Defend: Towards Chinese Toxicity Attacks via Implicit Enhancement and Obfuscation Rewriting

Jingyi Kang, Junyu Lu, Bo Xu, Hongbo Wang +3 more

Published

May 21, 2026

Citations

0

Trust level

Moderate

Usefulness score

67/100 (Medium)

Extraction confidence

70% (Moderate)

Derived from extracted protocol signals and abstract evidence.

Rater population

Not reported

Signals refreshed

May 21, 2026

Open arXiv Find implementation Shortlist (0) RSS feed

Tags

General Human Eval Red Team

Should you rely on this paper?

This paper has useful evaluation signal, but protocol completeness is partial; pair it with related papers before deciding implementation strategy.

Use this for comparison and orientation, not as your only source.

Best use

Secondary protocol comparison source

Use if you need

A secondary eval reference to pair with stronger protocol papers.

What to verify

Validate the evaluation procedure and quality controls in the full paper before operational use.

Main weakness

No major weakness surfaced.

Human feedback signal
Detected
From extracted signals
Evaluation signal
Detected
Eval setup described
Usefulness for eval research
67/100
Moderate-confidence candidate

Useful as a secondary reference; validate protocol details against neighboring papers.

If you are doing eval pipeline work, start here

Human Eval Papers LLM-as-Judge Papers Pairwise Preference Papers Tool-Use Evaluation Search

Abstract

Large language models (LLMs) require robust toxicity evaluation beyond explicit wording. This setting remains underexplored in Chinese, where toxicity may combine semantic indirectness with surface obfuscation. We introduce Chinese Implicit Toxicity Attack (CITA), a controlled red-team evaluation and defense-data generation framework, not a deployable evasion tool. CITA uses three stages: (i) Harmful Intent Learning, (ii) Implicit Toxicity Enhancement, and (iii) Obfuscation Variant Rewriting, to preserve harmful intent, increase implicitness, and add controlled surface variants. On CITA-generated evaluation samples, the seven tested detectors exhibit substantial missed-detection risks, reaching an average ASR of 69.48%; human evaluation further confirms preserved harmfulness and increased implicitness/evasiveness. As a downstream defense application, we fine-tune a Chinese Implicit Toxicity Defense model (CITD) with CITA-generated red-team data, showing that such data can improve robustness through additional training.

What we could verify

These are the protocol signals we could actually recover from the available paper metadata. Use them to decide whether this paper is worth deeper reading.

Human Feedback Types

strong

Red Team

Directly usable for protocol triage.

"Large language models (LLMs) require robust toxicity evaluation beyond explicit wording."

Evaluation Modes

strong

Human Eval

Includes extracted eval setup.

"Large language models (LLMs) require robust toxicity evaluation beyond explicit wording."

Quality Controls

missing

Not reported

No explicit QC controls found.

"Large language models (LLMs) require robust toxicity evaluation beyond explicit wording."

Benchmarks / Datasets

missing

Not extracted

No benchmark anchors detected.

"Large language models (LLMs) require robust toxicity evaluation beyond explicit wording."

Reported Metrics

strong

Toxicity, Jailbreak success rate

Useful for evaluation criteria comparison.

"Large language models (LLMs) require robust toxicity evaluation beyond explicit wording."

Benchmarks and datasets

No benchmark or dataset names were extracted from the available abstract.

Reported metrics

toxicityjailbreak success rate
Human feedback details
Uses human feedback
Yes
Feedback types
Red Team
Rater population
Not reported
Expertise required
General
Evaluation details
Evaluation modes
Human Eval
Agentic eval
None
Quality controls
Not reported
Evidence quality
Moderate
Use this page as
Secondary protocol comparison source

Research brief

Metadata summary

Large language models (LLMs) require robust toxicity evaluation beyond explicit wording.

Based on abstract + metadata only. Check the source paper before making high-confidence protocol decisions.

Key takeaways

  • Large language models (LLMs) require robust toxicity evaluation beyond explicit wording.
  • This setting remains underexplored in Chinese, where toxicity may combine semantic indirectness with surface obfuscation.
  • We introduce Chinese Implicit Toxicity Attack (CITA), a controlled red-team evaluation and defense-data generation framework, not a deployable evasion tool.

Researcher actions

  • Compare this paper against nearby papers in the same arXiv category before using it for protocol decisions.
  • Validate inferred eval signals (Human evaluation) against the full paper.
  • Use related-paper links to find stronger protocol-specific references.

Caveats

  • Generated from abstract + metadata only; no PDF parsing.
  • Signals below are heuristic and may miss details reported outside the abstract.

Recommended queries

Red teaming evaluation

Contribution summary

  • Large language models (LLMs) require robust toxicity evaluation beyond explicit wording.
  • We introduce Chinese Implicit Toxicity Attack (CITA), a controlled red-team evaluation and defense-data generation framework, not a deployable evasion tool.
  • On CITA-generated evaluation samples, the seven tested detectors exhibit substantial missed-detection risks, reaching an average ASR of 69.48%; human evaluation further confirms preserved harmfulness and increased implicitness/evasiveness.

Why it matters for eval

  • We introduce Chinese Implicit Toxicity Attack (CITA), a controlled red-team evaluation and defense-data generation framework, not a deployable evasion tool.
  • On CITA-generated evaluation samples, the seven tested detectors exhibit substantial missed-detection risks, reaching an average ASR of 69.48%; human evaluation further confirms preserved harmfulness and increased implicitness/evasiveness.

Researcher checklist

  • Human feedback protocol is explicit

    Detected: Red Team

  • Evaluation mode is explicit

    Detected: Human Eval

  • Quality control reporting appears

    No calibration/adjudication/IAA control explicitly detected.

  • Benchmark or dataset anchors are present

    No benchmark/dataset anchor extracted from abstract.

  • Metric reporting is present

    Detected: toxicity, jailbreak success rate