# Mechanistic Interpretability of GPT-2: Lexical and Contextual Layers in Sentiment Analysis

arXiv: https://arxiv.org/abs/2512.06681 · HTML: https://mlsanity.com/paper/2512.06681/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Circuits and mechanistic analysis, one of 13 directions in Mechanistic interpretability.
- Builds on 9 of its 16 references that are in the map.
- First posted on arXiv 7 Dec 2025.

## Placed in

- Mechanistic interpretability / [Circuits and mechanistic analysis](https://mlsanity.com/mechinterp/circuits-and-mechanistic-analysis/)

## Builds on (most cited first, 9 of 9)

- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [BERT Rediscovers the Classical NLP Pipeline](https://mlsanity.com/paper/1905.05950/) (2019)
- [What Does BERT Look At? An Analysis of BERT's Attention](https://mlsanity.com/paper/1906.04341/) (2019)
- [A Primer in BERTology: What we know about how BERT works](https://mlsanity.com/paper/2002.12327/) (2020)
- [Designing and Interpreting Probes with Control Tasks](https://mlsanity.com/paper/1909.03368/) (2019)
- [How to use and interpret activation patching](https://mlsanity.com/paper/2404.15255/) (2024)
- [Adversarial Removal of Demographic Attributes from Text Data](https://mlsanity.com/paper/1808.06640/) (2018)
- [A Practical Review of Mechanistic Interpretability for Transformer-Based Language Models](https://mlsanity.com/paper/2407.02646/) (2024)
- [Is This the Subspace You Are Looking for? An Interpretability Illusion for Subspace Activation Patching](https://mlsanity.com/paper/2311.17030/) (2023)

## Abstract (arXiv)

We present a mechanistic interpretability study of GPT-2 that causally examines how sentiment information is processed across its transformer layers. Using systematic activation patching across all 12 layers, we test the hypothesized two-stage sentiment architecture comprising early lexical detection and mid-layer contextual integration. Our experiments confirm that early layers (0-3) act as lexical sentiment detectors, encoding stable, position specific polarity signals that are largely independent of context. However, all three contextual integration hypotheses: Middle Layer Concentration, Phenomenon Specificity, and Distributed Processing are falsified. Instead of mid-layer specialization, we find that contextual phenomena such as negation, sarcasm, domain shifts etc. are integrated primarily in late layers (8-11) through a unified, non-modular mechanism. These experimental findings provide causal evidence that GPT-2's sentiment computation differs from the predicted hierarchical pattern, highlighting the need for further empirical characterization of contextual integration in large language models.