MechELK: A Mechanistic Interpretability Framework for Eliciting Latent Knowledge in Large Language Models

Ji-jun Park; Soo-joon Choi; Jiwon Jeong; Taeyang Yoon; Ju-Wan Lee

cs.CLApr 7, 2026

MechELK: A Mechanistic Interpretability Framework for Eliciting Latent Knowledge in Large Language Models

Authors: Ji-jun Park, Soo-joon Choi, Jiwon Jeong, Taeyang Yoon, Ju-Wan Lee

Organizations: Dongguk University

Abstract

Large language models (LLMs) frequently encode factual and reasoning knowledge in their internal representations that is not faithfully reflected in their surface-level outputs -- a phenomenon known as latent knowledge. Existing approaches to eliciting latent knowledge, such as Contrastive Consistency Search (CCS), rely on contrastive activation patterns and struggle with complex multi-step reasoning tasks, while mechanistic interpretability tools have primarily been used to understand model behavior rather than to extract hidden knowledge. We present MechELK, a unified three-stage framework that bridges mechanistic interpretability and latent knowledge elicitation. MechELK operates through: (1) Locate -- using Sparse Autoencoder (SAE) feature analysis and activation patching to identify knowledge-bearing representations; (2) Verify -- employing causal probing to distinguish genuine latent knowledge from spurious correlations; and (3) Elicit -- applying representation engineering to surface hidden knowledge without modifying model weights. Evaluated on TruthfulQA, a curated Deceptive Alignment benchmark, and the Quirky LM dataset, MechELK achieves an average elicitation accuracy of 84.7%, outperforming CCS by 6.2% and direct linear probing by 9.1%. Crucially, MechELK successfully identifies latent knowledge in 78.3% of cases where the model's surface output is incorrect or evasive, demonstrating its utility for AI safety applications including deceptive alignment detection.

Figures & tables

Method	TruthfulQA		Quirky LM		DAB
Method	Llama-8B	Llama-70B	Llama-8B	Llama-70B	Llama-8B	Mistral-7B
Direct Probing	68.4	72.1	71.3	75.6	62.1	60.8
CCS	74.2	78.5	76.8	81.2	67.3	65.9
RepE	72.8	76.3	74.1	79.4	70.2	68.7
SAE-Probe	75.6	79.8	77.4	82.1	69.8	67.3
Act. Patching	73.1	77.4	75.9	80.3	68.4	66.1

Table 1: Main results: Elicitation Accuracy (%) on three benchmarks. Best results are bold ; second-best are underlined .

\Delta

denotes improvement over CCS.

Method	DR (%) $\uparrow$	FPR (%) $\downarrow$	CS $\uparrow$	EA (%) $\uparrow$	Latency (s)
Direct Probing	81.2	28.4	0.61	68.4	0.1
CCS	83.7	22.1	0.68	74.2	8.7
RepE	79.4	19.8	0.72	72.8	0.3
SAE-Probe	85.3	18.6	0.74	75.6	1.2
Act. Patching	82.1	21.3	0.69	73.1	4.1
MechELK	91.4	12.7	0.89	82.3	3.2

Table 2: Detection Rate (DR), False Positive Rate (FPR), and Consistency Score (CS) on Llama-3-8B. Lower FPR and higher DR/CS are better.

Configuration	EA (%)	DR (%)	FPR (%)	CS
Full MechELK	82.3	91.4	12.7	0.89
w/o Verify (CKS filtering)	76.1	88.2	24.3	0.74
w/o SAE (use raw activations)	77.4	85.6	19.8	0.71
w/o Layer Selection (use last layer)	74.8	83.1	22.6	0.68
w/o Feature Differential (use $\mathbf{f}_{\ell}(x_{y^{*}})$ only)	78.2	87.4	21.1	0.76
w/o Elicit (use Verify output as classifier)	79.6	91.4	12.7	0.89

Table 3: Ablation study on TruthfulQA (Llama-3-8B). Each row removes one component from the full MechELK pipeline.

Figure 1: Distribution of optimal knowledge layers

\ell^{*}

selected by MechELK across TruthfulQA (blue) and DAB (orange) queries on Llama-3-8B. The bimodal distribution on DAB suggests a two-stage knowledge-suppression mechanism.

Figure 2: Effect of CKS threshold

\tau

on elicitation accuracy (EA), detection rate (DR), and false positive rate (FPR) on TruthfulQA. The optimal threshold

\tau^{*}=0.15

is stable across models.

Figure 3: Elicitation accuracy as a function of intervention strength

\lambda

on three benchmarks. The optimal strength

\lambda^{*}\approx 1.2

is consistent across datasets, suggesting a universal elicitation regime.

Figure 4: Distribution of pairwise cosine similarities between knowledge directions for paraphrased queries. MechELK (mean=0.89) substantially outperforms CCS (0.68) and direct probing (0.61).

Figure 5: Elicitation accuracy vs. model size for MechELK and CCS. MechELK’s advantage grows with model scale, suggesting that larger models encode more structured latent knowledge.

Figure 6: Case studies from the Deceptive Alignment Benchmark. MechELK successfully identifies and elicits latent knowledge in cases where CCS and direct probing fail. CKS values above

\tau^{*}=0.15

indicate detected latent knowledge.

Failure Mode	Frequency (%)	Avg. CKS
Knowledge fragmentation (multi-layer)	42.3	0.08
SAE reconstruction error	31.1	0.11
Intervention disruption	15.7	0.19
Genuine absence of knowledge	10.9	0.03

Table 4: Failure mode analysis for MechELK on TruthfulQA (Llama-3-8B).

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Method	Model	TruthfulQA	Quirky LM	DAB
MechELK	Llama-3-8B	$82.3\pm 0.4$	$83.1\pm 0.6$	$81.2\pm 0.8$
	Llama-3-70B	$86.7\pm 0.3$	$87.4\pm 0.4$	$85.9\pm 0.5$
	Mistral-7B	$80.1\pm 0.5$	$81.8\pm 0.7$	$79.6\pm 0.9$
CCS	Llama-3-8B	$74.2\pm 0.8$	$76.8\pm 1.1$	$67.3\pm 1.4$
	Llama-3-70B	$78.5\pm 0.6$	$81.2\pm 0.9$	$72.1\pm 1.2$
	Mistral-7B	$72.1\pm 0.9$	$74.3\pm 1.2$	$65.9\pm 1.5$

Appendix

Table 5: Full results with standard deviations (3 seeds).

LLM Interpretability

Large language model (LLM) interpretability studies how internal representations and computations produce outputs, and how they can be explained or controlled.

See all 778 papers

Mechanistically Eliciting Latent Behaviors in Language Models
Jun 28, 2026Andrew Mack, Nina Panickssery, Alexander Matt TurnerLLM Alignment Language Model Safety Evaluation
†Principles of Intelligence · ‡Anthropic · §Independent
Towards Robust Model Organisms for Eliciting Latent Knowledge
Oct 19, 2026Lennart Finke, Hardik Bhatnagar, Euodia Dodd +4AI Safety Language Model Safety Evaluation
ETH Zürich MATS · University of Tübingen Tübingen AI Center MATS · Imperial College London MATS +2
MechLens: Late Crystallization of Factual Knowledge Explains Intervention Effectiveness in Language Models
Jun 6, 2026Xueping GaoMemorization in Language Models LLM Interpretability
Alibaba Cloud

Jun 28, 2026J/K move · Enter open · S save

Mechanistically Eliciting Latent Behaviors in Language Models

Andrew Mack, Nina Panickssery, Alexander Matt Turner

†Principles of Intelligence · ‡Anthropic · §Independent

LLM Alignment Language Model Safety Evaluation LLM Interpretability Low-Rank Adaptation Causal Interventions in Language Models

Open paper AskPDF arXiv

We aim to discover diverse, generalizable perturbations of LLM internals that can surface hidden behavioral modes. Such perturbations could help reshape model behavior and systematically evaluate potential risks. We introduce Causal Perturbative Elicitation (CPE), an unsupervised method for discovering interpretable low-rank adapters (LoRAs) that can elicit these latent behaviors. CPE decomposes the computations of a deep transformer slice using a heuristic tensor-decomposition-based algorithm. CPE exhibits remarkable data efficiency, learning a large number of interpretable LoRAs from a single example. Even though CPE is unsupervised, we find that in some cases it can be competitive with supervised elicitation methods via brute-force enumerative search over weight space. For instance, CPE performs similarly to matched-wall-clock-time GRPO on the Countdown task for Qwen3-8B (85% vs 87%), demonstrating that CPE can efficiently elicit complex multi-token behaviors. Since CPE is unsupervised, it can also surface hidden failure modes, such as sandbagging, restoring 85% of locked BigCodeBench performance on a password-locked version of Llama3-70B introduced by Taylor et al. (2025). Additionally, since CPE explores behaviors in weight-space rather than token-space it can potentially ameliorate exploration hacking, a misalignment failure which may arise in sufficiently self-aware AI models (Ngo, 2022). In fact, we find that CPE virtually eliminates alignment-faking (Greenblatt et al., 2024) behavior in a Llama3-70B-based model organism developed by Hughes et al. (2025). Finally, we find that CPE can be used to initialize GPT-OSS-20B in an aligned basin when running GRPO on an environment prone to reward-hacking. By providing a data-efficient method to systematically explore the space of latent model behaviors, CPE yields a powerful tool for aligning AI systems and evaluating their safety.

Abstract

LLM Interpretability

Explore similar work

Mechanistically Eliciting Latent Behaviors in Language Models

Towards Robust Model Organisms for Eliciting Latent Knowledge

MechLens: Late Crystallization of Factual Knowledge Explains Intervention Effectiveness in Language Models