mlsanityData to 7 Oct 2026

Beyond Hidden-Layer Manipulation: Semantically-Aware Logit Interventions for Debiasing LLMs

arXiv 2510.23650 ↗First posted 25 Oct 2025cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

We proposed Static and Dynamic -- two zero-shot logits-layer debiasing methods. Dynamic reduces bias by up to 70% with minimal fluency loss. Logits intervention outperforms hidden-layer approaches. We show semantic-aware logits intervention is stable and effective for debiasing aligned LLMs.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 19 of its 21 references that are in the map. First posted on arXiv 25 Oct 2025.

Builds on

19 of 21 references.

References · 19

  1. 2024The Llama 3 Herd of Modelscited by 14,192
  2. 2023Qwen Technical Reportcited by 3,273
  3. 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,565
  4. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  5. 2020CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Modelscited by 703
  6. 2021BBQ: A Hand-Built Bias Benchmark for Question Answeringcited by 683
  7. 2020Intrinsic Dimensionality Explains the Effectiveness of Language Model Fine-Tuningcited by 658
  8. 2018Gender Bias in Coreference Resolution: Evaluation and Debiasing Methodscited by 657
  9. 2022Contrastive Decoding: Open-ended Text Generation as Optimizationcited by 598
  10. 2023DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Modelscited by 367
  11. 2024Do Llamas Work in English? On the Latent Language of Multilingual Transformerscited by 275
  12. 2023Aligning Large Language Models with Human Preferences through Representation Engineeringcited by 71
7 more
  1. 2024Measuring Implicit Bias in Explicitly Unbiased Large Language Modelscited by 49
  2. 2024UniBias: Unveiling and Mitigating LLM Bias through Internal Attention and FFN Manipulationcited by 30
  3. 2025Taxonomy, Opportunities, and Challenges of Representation Engineering for Large Language Modelscited by 29
  4. 2025The Mirage of Model Editing: Revisiting Evaluation in the Wildcited by 24
  5. 2024Analysing the Residual Stream of Language Models Under Knowledge Conflictscited by 15
  6. 2023Knowledgeable In-Context Tuning: Exploring and Exploiting Factual Knowledge for In-Context Learningcited by 6
  7. 2025K-Edit: Language Model Editing with Contextual Knowledge Awarenesscited by 4

Built on by

Not cited yet.