# Correctness-Optimized Residual Activation Lens (CORAL): Transferrable and Calibration-Aware Inference-Time Steering

arXiv: https://arxiv.org/abs/2602.06022 · HTML: https://mlsanity.com/paper/2602.06022/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 33 references that are in the map.
- First posted on arXiv 5 Feb 2026.

## Builds on (most cited first, 23 of 23)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge](https://mlsanity.com/paper/1811.00937/) (2018)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [RACE: Large-scale ReAding Comprehension Dataset From Examinations](https://mlsanity.com/paper/1704.04683/) (2017)
- [Towards Automated Circuit Discovery for Mechanistic Interpretability](https://mlsanity.com/paper/2304.14997/) (2023)
- [MathQA: Towards Interpretable Math Word Problem Solving with Operation-Based Formalisms](https://mlsanity.com/paper/1905.13319/) (2019)
- [Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space](https://mlsanity.com/paper/2203.14680/) (2022)
- [How Can We Know When Language Models Know? On the Calibration of Language Models for Question Answering](https://mlsanity.com/paper/2012.00955/) (2020)
- [Taming Overconfidence in LLMs: Reward Calibration in RLHF](https://mlsanity.com/paper/2410.09724/) (2024)
- [Confidence Regulation Neurons in Language Models](https://mlsanity.com/paper/2406.16254/) (2024)
- [pyvene: A Library for Understanding and Improving PyTorch Models via Interventions](https://mlsanity.com/paper/2403.07809/) (2024)
- [No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes](https://mlsanity.com/paper/2509.10625/) (2025)
- [Careful Selection of Knowledge to solve Open Book Question Answering](https://mlsanity.com/paper/1907.10738/) (2019)
- [Calibrating LLM Confidence by Probing Perturbed Representation Stability](https://mlsanity.com/paper/2505.21772/) (2025)
- [SteerConf: Steering LLMs for Confidence Elicitation](https://mlsanity.com/paper/2503.02863/) (2025)
- [CorrSteer: Generation-Time LLM Steering via Correlated Sparse Autoencoder Features](https://mlsanity.com/paper/2508.12535/) (2025)
- [Calibration Across Layers: Understanding Calibration Evolution in LLMs](https://mlsanity.com/paper/2511.00280/) (2025)

## Built on by (most cited first, 2 of 2)

- [Conceptors for Semantic Steering](https://mlsanity.com/paper/2605.04980/) (2026)
- [Contrastive Conceptor Activation Steering (COAST): Unlocking Vision-Language-Action Models through Hidden States](https://mlsanity.com/paper/2605.17144/) (2026)

## Abstract (arXiv)

Large language models (LLMs) exhibit persistent miscalibration, especially after instruction tuning and preference alignment. Modified training objectives can improve calibration, but retraining is expensive. Inference-time steering offers a lightweight alternative, yet most existing methods optimize proxies for correctness rather than correctness itself. We introduce CORAL (Correctness-Optimized Residual Activation Lens), a regularized inference-time steering method that captures distributed correctness signals from model internal activations using weight-decay MLP probes. We evaluate CORAL across three 7B-parameter models and find that it consistently improves accuracy by 10% and expected calibration error (ECE) by 50% on average. We additionally demonstrate that these gains transfer without retraining to the complete published test sets of four held-out benchmarks (ARC-Challenge, HellaSwag, Math-MC, OpenBookQA), averaging 14% accuracy improvements and 49% ECE improvements. Our results support the hypothesis that distributed information in model internals can be extracted using regularized probes when individual neurons are insufficient. CORAL thus provides a compute-efficient, transferable, and calibration-aware approach to improve MCQA performance during inference.