mlsanityData to 14 Oct 2026

Transcoders for Investigating Deception in Language Models

arXiv 2607.14791 ↗First posted 16 Jul 2026cs.AIcited by 0

On the map: Mechanistic interpretability → Probing deception, bias and beliefs

Abstract (arXiv)

Transcoders have recently emerged as a promising approach for mechanistic interpretability (MI), enabling circuit-level analysis of model behaviour. In this paper, we investigate the use of transcoders to analyse deceptive behaviour in language models, a behaviour that poses a safety and security risk. Using a Qwen3-4B model with pre-trained transcoders, specifically per-layer transcoders (PLTs), we construct attribution graphs that capture feature activations and inter-feature dependencies, allowing circuit-level analysis of deception. Through feature steering and circuit analysis, we identified a dictionary of deception-related features and show that these features exert a stronger influence on deceptive outputs, as they produce predictable shifts between deceptive and non-deceptive responses. These findings suggest that deception emerges from internal model mechanisms and highlight the potential of transcoders for behavioural monitoring and early detection of security vulnerabilities related to malicious behaviours in language models.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Probing deception, bias and beliefs, one of 13 directions in Mechanistic interpretability. Builds on 11 of its 20 references that are in the map. First posted on arXiv 16 Jul 2026.

Builds on

11 of 20 references.

Closest ancestors · 2

Same direction, newest first

  1. 2025When Thinking LLMs Lie: Unveiling the Strategic Deception in Representations of Reasoning Models5 Jun
  2. 2025Detecting Strategic Deception Using Linear Probes5 Feb

Other references · 9

  1. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
  2. 2024Transcoders Find Interpretable LLM Feature Circuitscited by 200
  3. 2024Truth is Universal: Robust Detection of Lies in LLMscited by 79
  4. 2025Transcoders Beat Sparse Autoencoders for Interpretabilitycited by 25
  5. 2023Exploiting Large Language Models (LLMs) through Deception Techniques and Persuasion Principlescited by 17
  6. 2024Detecting and Understanding Vulnerabilities in Language Models via Mechanistic Interpretabilitycited by 13
  7. 2025Revisiting End-To-End Sparse Autoencoder Training: A Short Finetune Is All You Needcited by 4
  8. 2024Banal Deception Human-AI Ecosystems: A Study of People's Perceptions of LLM-generated Deceptive Behaviourcited by 2
  9. 2025Evaluating the Evaluators: Trust in Adversarial Robustness Testscited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Debate with Images: Detecting Deceptive Behaviors in Multimodal Large Language Models
  2. 2026Beyond Liars' Bench: The Impact of Lie Typology, Depth, and Sparsity on Deception Detection in LLMs
  3. 2025Evaluating & Reducing Deceptive Dialogue From Language Models with Multi-turn RL
  4. 2026Decoding Hidden Deception in Reasoning LLMs: Activation Explainers for Deception Auditing
  5. 2025When Truthful Representations Flip Under Deceptive Instructions?
  6. 2026DeceptGuard :A Constitutional Oversight Framework For Detecting Deception in LLM Agents
  7. 2026Fuzzing Large Language Models to Elicit Hidden Behaviours
  8. 2026"Did you lie?" Evaluating Lie Detectors across Model Scale and Belief-Verified Model Organisms