Abstract
Activation verbalization methods such as Activation Oracle and Natural Language Autoencoders decode hidden representations of large language models into human-readable natural language. However, existing methods can produce incomplete or hallucinated descriptions, making their activation verbalizations difficult to trust and use reliably in practice. To this end, we introduce AVPO, a two-stage framework that first reconstructs source text from a hidden activation and then evaluates the resulting text with a separate frozen question-answering model, yielding an explicit and inspectable intermediate readout. We further optimize the inverter with direct preference optimization (DPO), using rewards that capture both semantic recoverability and lexical fidelity. Across six text families, AVPO improves gist- and detail-level information recovery over the strongest baseline by up to 17.1 and 9.3 percentage points, respectively. Crucially, the gains arise from preference optimization rather than fine-tuning on selected reconstructions alone, enabling compact cross-model inverters to surpass donor-matched question-conditioned verbalizers while improving both semantic recoverability and lexical fidelity. Moreover, out-of-distribution case study shows that AVPO better recovers high-level semantics while fabricating fewer details.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Zhao, H., Hei, Z., Shi, W., Deng, H., Zou, N., & Du, M. (2026). Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation. https://omanscience.com/en/articles/faithful-activation-verbalization-reducing-hallucinations-in-llm-representation-interpretation
MLA 9
Zhao, Haiyan, et al. "Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation." https://omanscience.com/en/articles/faithful-activation-verbalization-reducing-hallucinations-in-llm-representation-interpretation.
Chicago (author–date)
Zhao, Haiyan, Zirui Hei, Wei Shi, Huiqi Deng, Na Zou, and Mengnan Du. 2026. "Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation." https://omanscience.com/en/articles/faithful-activation-verbalization-reducing-hallucinations-in-llm-representation-interpretation.
Harvard
Zhao, H., Hei, Z., Shi, W., Deng, H., Zou, N. and Du, M. (2026) 'Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation', Available at: https://omanscience.com/en/articles/faithful-activation-verbalization-reducing-hallucinations-in-llm-representation-interpretation.
Vancouver
Zhao H, Hei Z, Shi W, Deng H, Zou N, Du M. Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation. https://omanscience.com/en/articles/faithful-activation-verbalization-reducing-hallucinations-in-llm-representation-interpretation
IEEE
H. Zhao, Z. Hei, W. Shi, H. Deng, N. Zou, and M. Du, "Faithful Activation Verbalization: Reducing Hallucinations in LLM Representation Interpretation," https://omanscience.com/en/articles/faithful-activation-verbalization-reducing-hallucinations-in-llm-representation-interpretation.