# Epistemic Goggles: A Pretrained Module that Induces an Epistemic Frame via Gradient Editing

arXiv: https://arxiv.org/abs/2607.01690 · HTML: https://mlsanity.com/paper/2607.01690/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 27 of its 33 references that are in the map.
- First posted on arXiv 2 Jul 2026.

## Builds on (most cited first, 27 of 27)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [A General Language Assistant as a Laboratory for Alignment](https://mlsanity.com/paper/2112.00861/) (2021)
- [Tulu 3: Pushing Frontiers in Open Language Model Post-Training](https://mlsanity.com/paper/2411.15124/) (2024)
- [OpenAssistant Conversations -- Democratizing Large Language Model Alignment](https://mlsanity.com/paper/2304.07327/) (2023)
- [Learning to learn by gradient descent by gradient descent](https://mlsanity.com/paper/1606.04474/) (2016)
- [Fast Model Editing at Scale](https://mlsanity.com/paper/2110.11309/) (2021)
- [Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research](https://mlsanity.com/paper/2402.00159/) (2024)
- [Measuring short-form factuality in large language models](https://mlsanity.com/paper/2411.04368/) (2024)
- [Gradient-based Hyperparameter Optimization through Reversible Learning](https://mlsanity.com/paper/1502.03492/) (2015)
- [Learning by Distilling Context](https://mlsanity.com/paper/2209.15189/) (2022)
- [Massive Editing for Large Language Models via Meta Learning](https://mlsanity.com/paper/2311.04661/) (2023)
- [Personas as a Way to Model Truthfulness in Language Models](https://mlsanity.com/paper/2310.18168/) (2023)
- [Steering Evaluation-Aware Language Models to Act Like They Are Deployed](https://mlsanity.com/paper/2510.20487/) (2025)
- [Reinforced Lifelong Editing for Language Models](https://mlsanity.com/paper/2502.05759/) (2025)
- [Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment](https://mlsanity.com/paper/2601.10160/) (2026)
- [Believe It or Not: How Deeply do LLMs Believe Implanted Facts?](https://mlsanity.com/paper/2510.17941/) (2025)
- [Emergent Misalignment is Easy, Narrow Misalignment is Hard](https://mlsanity.com/paper/2602.07852/) (2026)
- [Model Spec Midtraining: Improving How Alignment Training Generalizes](https://mlsanity.com/paper/2605.02087/) (2026)
- [Implicit meta-learning may lead language models to trust more reliable sources](https://mlsanity.com/paper/2310.15047/) (2023)
- [Value Drifts: Tracing Value Alignment During LLM Post-Training](https://mlsanity.com/paper/2510.26707/) (2025)
- [Negation Neglect: When models fail to learn negations in training](https://mlsanity.com/paper/2605.13829/) (2026)
- [Just as Humans Need Vaccines, So Do Models: Model Immunization to Combat Falsehoods](https://mlsanity.com/paper/2505.17870/) (2025)
- [PropMEND: Hypernetworks for Knowledge Propagation in LLMs](https://mlsanity.com/paper/2506.08920/) (2025)
- [EMSEdit: Efficient Multi-Step Meta-Learning-based Model Editing](https://mlsanity.com/paper/2508.04012/) (2025)
- [Hierarchical Orthogonal Residual Spread for Precise Massive Editing in Large Language Models](https://mlsanity.com/paper/2601.11441/) (2026)

## Built on by (most cited first, 1 of 1)

- [Semantic Overlays: Mitigating Prompt Injection with Annotations Beyond Tokens and Steering Vectors](https://mlsanity.com/paper/2608.23873/) (2026)

## Abstract (arXiv)

Finetuning a language model on documents that are explicitly annotated as fictional results in a model that still actually believes the documents' core claims, an effect known as Negation Neglect. In our evaluations, models trained on documents prefixed and suffixed with such annotations correctly identify the relevant claims as fictional only about 9% of the time. To address this, we introduce Goggles, a learned module that intervenes on the finetuning gradient rather than the data. During supervised finetuning, a Goggles module edits the gradients an LLM LoRA receives, imparting a chosen epistemic frame (the stance the model takes toward the nature of what it reads) to whatever the documents teach. A Goggles instance is trained once for a given base model, frame, and LoRA configuration, then applied frozen to documents it was never trained on. Trained through Goggles on those same documents, now carrying no fictional annotation, the model flags the content as fictional roughly 91% of the time, while preserving capability (GPQA and TruthfulQA match or exceed baseline). The same architecture supports other frames: a Goggles instance can be trained to treat documents as "part of an AI safety evaluation by Redwood Research" rather than simply as fiction. The imparted frame persists under continued finetuning that pushes back toward the claim, where prior interventions revert. Goggles suggests a path toward training language models on known-misaligned data without absorbing the behaviors that data demonstrates.