MétaCan
Menu
Back to cohort

Evaluating Toxicity Understanding of LLM Agents

2025· article· W7125580424 on OpenAlexaff
Ramaravind Kommiya Mothilal, Shion Guha

Bibliographic record

Venuenot available
Typearticle
Language
FieldComputer Science
TopicHate Speech and Cyberbullying Detection
Canadian institutionsUniversity of Toronto
Fundersnot available
KeywordsToxicityPerceptionTrustworthinessOrder (exchange)Action (physics)Risk assessment

Abstract

fetched live from OpenAlex

Research on toxicity in LLMs has largely focused on detection tasks, such as identifying hate speech or stereotyping in texts. Recently, these tasks have increasingly been embedded in agentic workflows, where LLMs autonomously query external APIs and reason over results before responding. This shift promotes the perception that LLMs exhibit an “understanding” of toxicity, yet how such understanding can be meaningfully interpreted by humans remains unclear. In this position paper, we first unpack this oversight by highlighting the fundamental gaps in current literature and then propose a framework for evaluating toxicity understanding of agentic LLMs. Overall, this short paper aims to shift the discourse from improving toxicity detection in LLMs to evaluating how LLMs understand toxicity in order to enhance their trustworthiness in downstream tasks

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.013
metaresearch head score (Gemma)0.089
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Observational · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.013
Threshold uncertainty score0.070

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0130.089
Meta-epidemiology (narrow)0.0010.000
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0020.001
Science and technology studies0.0010.002
Scholarly communication0.0040.007
Open science0.0010.005
Research integrity0.0020.002
Insufficient payload (model declined to judge)0.0030.001

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.134
GPT teacher head0.367
Teacher spread0.233 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designObservational
Domainnot available
GenreEmpirical

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2025
Admission routes1
Has abstractyes

Explore more

Same topicHate Speech and Cyberbullying DetectionFrench-language works237,207