# Pragmatic Inference Chain (PIC) Improving LLMs' Reasoning of Authentic Implicit Toxic Language

arXiv: https://arxiv.org/abs/2503.01539 · HTML: https://mlsanity.com/paper/2503.01539/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning.
- Builds on 26 of its 65 references that are in the map.
- First posted on arXiv 3 Mar 2025.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)

## Builds on (most cited first, 26 of 26)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models](https://mlsanity.com/paper/2009.11462/) (2020)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [DeepSeek-V3 Technical Report](https://mlsanity.com/paper/2412.19437/) (2024)
- [BBQ: A Hand-Built Bias Benchmark for Question Answering](https://mlsanity.com/paper/2110.08193/) (2021)
- [To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning](https://mlsanity.com/paper/2409.12183/) (2024)
- [Having Beer after Prayer? Measuring Cultural Bias in Large Language Models](https://mlsanity.com/paper/2305.14456/) (2023)
- [Think Twice: Perspective-Taking Improves Large Language Models' Theory-of-Mind Capabilities](https://mlsanity.com/paper/2311.10227/) (2023)
- [Are DeepSeek R1 And Other Reasoning Models More Faithful?](https://mlsanity.com/paper/2501.08156/) (2025)
- [PUB: A Pragmatics Understanding Benchmark for Assessing LLMs' Pragmatics Capabilities](https://mlsanity.com/paper/2401.07078/) (2024)
- [The Goldilocks of Pragmatic Understanding: Fine-Tuning Strategy Matters for Implicature Resolution by LLMs](https://mlsanity.com/paper/2210.14986/) (2022)
- [Controllable Text Generation for Large Language Models: A Survey](https://mlsanity.com/paper/2408.12599/) (2024)
- [ALONE: A Dataset for Toxic Behavior among Adolescents on Twitter](https://mlsanity.com/paper/2008.06465/) (2020)
- [Walking in Others' Shoes: How Perspective-Taking Guides Large Language Models in Reducing Toxicity and Bias](https://mlsanity.com/paper/2407.15366/) (2024)
- [AdaMCoT: Rethinking Cross-Lingual Factual Reasoning through Adaptive Multilingual Chain-of-Thought](https://mlsanity.com/paper/2501.16154/) (2025)
- [CoAT: Chain-of-Associated-Thoughts Framework for Enhancing Large Language Models Reasoning](https://mlsanity.com/paper/2502.02390/) (2025)
- [How Are LLMs Mitigating Stereotyping Harms? Learning from Search Engine Studies](https://mlsanity.com/paper/2407.11733/) (2024)
- [Constrained Reasoning Chains for Enhancing Theory-of-Mind in Large Language Models](https://mlsanity.com/paper/2409.13490/) (2024)
- [MyGO Multiplex CoT: A Method for Self-Reflection in Large Language Models via Double Chain of Thought Thinking](https://mlsanity.com/paper/2501.13117/) (2025)
- [DiPT: Enhancing LLM reasoning through diversified perspective-taking](https://mlsanity.com/paper/2409.06241/) (2024)
- [CDW-CoT: Clustered Distance-Weighted Chain-of-Thoughts Reasoning](https://mlsanity.com/paper/2501.12226/) (2025)
- [Chain-of-MetaWriting: Linguistic and Textual Analysis of How Small Language Models Write Young Students Texts](https://mlsanity.com/paper/2412.14986/) (2024)
- [DualCoTs: Dual Chain-of-Thoughts Prompting for Sentiment Lexicon Expansion of Idioms](https://mlsanity.com/paper/2409.17588/) (2024)
- [Chain of Alignment: Integrating Public Will with Expert Intelligence for Language Model Alignment](https://mlsanity.com/paper/2411.10534/) (2024)

## Built on by (most cited first, 8 of 8)

- [Discourse Heuristics For Paradoxically Moral Self-Correction](https://mlsanity.com/paper/2507.00985/) (2025)
- [Argument-Based Consistency in Toxicity Explanations of LLMs](https://mlsanity.com/paper/2506.19113/) (2025)
- [Diagnosing the Performance Trade-off in Moral Alignment: A Case Study on Gender Stereotypes](https://mlsanity.com/paper/2509.21456/) (2025)
- [From Training to Generalization: Improving Moral Reasoning Through Pragmatic Inference](https://mlsanity.com/paper/2509.24102/) (2025)
- [Can Large Language Models Handle Discourse Particles? A Case Study of Colloquial Malay](https://mlsanity.com/paper/2605.28782/) (2026)
- [Learning to Diagnose and Correct Moral Errors: Beyond Shallow Heuristics in Moral Alignment](https://mlsanity.com/paper/2601.03079/) (2026)
- [Towards High-Level Semantic Intelligence](https://mlsanity.com/paper/2607.24082/) (2026)
- [Pragmatic Attack Surface: Vulnerabilities of Implicit Context in Large Language Models](https://mlsanity.com/paper/2608.09551/) (2026)

## Abstract (arXiv)

The rapid development of large language models (LLMs) gives rise to ethical concerns about their performance, while opening new avenues for developing toxic language detection techniques. However, LLMs' unethical output and their capability of detecting toxicity have primarily been tested on language data that do not demand complex meaning inference, such as the biased associations of 'he' with programmer and 'she' with household. Nowadays toxic language adopts a much more creative range of implicit forms, thanks to advanced censorship. In this study, we collect authentic toxic interactions that evade online censorship and that are verified by human annotators as inference-intensive. To evaluate and improve LLMs' reasoning of the authentic implicit toxic language, we propose a new prompting method, Pragmatic Inference Chain (PIC), drawn on interdisciplinary findings from cognitive science and linguistics. The PIC prompting significantly improves the success rate of GPT-4o, Llama-3.1-70B-Instruct, DeepSeek-v2.5, and DeepSeek-v3 in identifying implicit toxic language, compared to five baseline prompts, such as CoT and rule-based baselines. In addition, it also facilitates the models to produce more explicit and coherent reasoning processes, hence can potentially be generalized to other inference-intensive tasks, e.g., understanding humour and metaphors.