arXiv preprint 2611.01710

Find the risk.
Trace it back.

CST: Lightweight Risk Triage with Source-Grounded Evidence for Web Discussions

CST helps moderators decide where to look, what to read, and how to check the evidence in the original discussion. The pipeline runs on a CPU.

Zhihua Wang1,2Yifeng Gao1,†Yunhao Feng3Rui Pan1Baiheng Zhao1Zixing Chen1Nan Li1

1 Fudan University · 2 University of Shanghai for Science and Technology · 3 Alibaba Group

† Corresponding author

Source trace
Source thread Hacker News
  1. m8reply to m1

    What kinds of user flags are there and why are they not public?

  2. m9reply to m8

    It's just the flagging you already have access to …

  3. m10reply to m9

    At some point you get the option to vouch for flagged posts as well.

  4. m11reply to m10

    >30 karma, same as flagging:

Evidence group

karma / flagging

Reply path
m8m9m10m11
Find discussions to review

Character features, sentence embeddings and reply links inform the queue.

Illustrative thread from Figures 3 and 8 · View paper figures

Three decisions in one review.

CST ranks discussions by risk, then separately selects and organizes evidence for review. Grouping leaves the risk scores unchanged.

Figure 3 Shared features, discussion ranking and source-linked review

What changes when you group the evidence?

Use the illustrative threads from Figures 1 and 2 to separate reply context, message selection, and presentation.

Figure 1

Same reply, different stance.

Change Alex’s message and watch what Sam’s “I agree.” endorses.

Alex

We should remove her from the group.

Sam · reply to Alex

I agree.

Sam supports the removal.

Compare the results.

Explore the paper’s evidence recall, classification performance, and end-to-end CPU cost.

Messages per discussion

136 discussions · 1,960 dangerous messages

  • Original order14.49%
  • Risk order31.12%
  • MMR32.14%
  • Group round-robin34.39%
Recall across budgets (%)
Selector 13510
Original order 1.38 4.13 7.14 14.49
Risk order 4.29 11.43 17.86 31.12
MMR 4.29 11.48 18.37 32.14
Group round-robin 4.29 12.55 20.05 34.39

Figure 4. Recall is measured across all dangerous messages in the selected cohort. Every discussion receives the selected budget.

Cite CST.

Read the full paper on arXiv or use the BibTeX entry in your bibliography.

BibTeX
@misc{wang2026cstlightweightrisktriage,
  title={CST: Lightweight Risk Triage with Source-Grounded Evidence for Web Discussions},
  author={Wang, Zhihua and Gao, Yifeng and Feng, Yunhao and Pan, Rui and
          Zhao, Baiheng and Chen, Zixing and Li, Nan},
  year={2026},
  eprint={2611.01710},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2611.01710}
}

Open image