Metraipolis › archive › dc_07c2e810f091

BiasTrace: Linking Reasoning Behaviours to Biased Outputs in LLMs — arXiv:2608.14161

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.14161
License (as archived)
CC0-1.0
Archived
2026-10-19T02:20:38.171Z
SHA-256 of the archived text
95ed46e2820d22dc73187b10ed93fb76f5aae4e6debd527b2f141b8b609c75db
Stable URL of this page
https://metraipolis.com/documento/dc_07c2e810f091

How to cite

BiasTrace: Linking Reasoning Behaviours to Biased Outputs in LLMs — arXiv:2608.14161 — arXiv (en), https://arxiv.org/abs/2608.14161 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_07c2e810f091 (Metraipolis archive dc_07c2e810f091, sha256 95ed46e2820d22dc73187b10ed93fb76f5aae4e6debd527b2f141b8b609c75db)
BibTeX
@misc{dc07c2e810f091,
  title = {BiasTrace: Linking Reasoning Behaviours to Biased Outputs in LLMs — arXiv:2608.14161},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.14161},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_07c2e810f091; sha256 95ed46e2820d22dc73187b10ed93fb76f5aae4e6debd527b2f141b8b609c75db},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.14161. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

LLMs exhibit social biases that can produce inaccurate and discriminatory inferences, posing risks in high-stakes applications. While prior work has made progress in measuring and mitigating bias, it largely focuses on final outputs of models, with limited understanding of the mechanisms that produce biased outcomes. Recent advances in LLM reasoning offers a new lens for investigating bias, yet the link between reasoning and bias remains poorly understood. Existing approaches focus primarily on final answer correctness or explicitly biased language, overlooking different behaviours in reasoning that can drive biased outcomes. We introduce BiasTrace, an annotation scheme for labelling reasoning behaviours in model-generated traces and linking them to biased outcomes. BiasTrace captures bias-specific behaviours (e.g., unsupported demographic assumptions) as well as general reasoning patterns that may implicitly contribute to bias (e.g. overthinking). We apply BiasTrace to reasoning traces in bias-sensitive contexts, scaled using validated LLM-as-a-judge methods, producing a large annotated dataset. Our analysis shows that biased outputs often stem from subtle reasoning behaviours rather than explicitly biased language, and that reasoning-level annotations improve bias detection. We further show that BiasTrace behaviours can be exploited for inference-time mitigation. These findings underscore the importance of examining a broader range of reasoning patterns to better understand bias in LLMs.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-17 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.14161 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-19T02:02:02.788Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-17?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).