# Knowing but Not Saying: Preventing Factual Access Failures in LLM SFT via Recall-Anchored Distillation

arXiv: https://arxiv.org/abs/2608.20794 · HTML: https://mlsanity.com/paper/2608.20794/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 40 of its 49 references that are in the map.
- First posted on arXiv 21 Aug 2026.

## Builds on (most cited first, 30 of 40)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Learning without Forgetting](https://mlsanity.com/paper/1606.09282/) (2016)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories](https://mlsanity.com/paper/2212.10511/) (2022)
- [An Empirical Study of Catastrophic Forgetting in Large Language Models During Continual Fine-tuning](https://mlsanity.com/paper/2308.08747/) (2023)
- [Instruction Tuning for Large Language Models: A Survey](https://mlsanity.com/paper/2308.10792/) (2023)
- [MedMCQA : A Large-scale Multi-Subject Multi-Choice Dataset for Medical domain Question Answering](https://mlsanity.com/paper/2203.14371/) (2022)
- [Towards Expert-Level Medical Question Answering with Large Language Models](https://mlsanity.com/paper/2305.09617/) (2023)
- [LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations](https://mlsanity.com/paper/2410.02707/) (2024)
- [Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?](https://mlsanity.com/paper/2405.05904/) (2024)
- [Recall and Learn: Fine-tuning Deep Pretrained Language Models with Less Forgetting](https://mlsanity.com/paper/2004.12651/) (2020)
- [Understanding Catastrophic Forgetting in Language Models via Implicit Inference](https://mlsanity.com/paper/2309.10105/) (2023)
- [Revisiting Catastrophic Forgetting in Large Language Model Tuning](https://mlsanity.com/paper/2406.04836/) (2024)
- [Unfamiliar Finetuning Examples Control How Language Models Hallucinate](https://mlsanity.com/paper/2403.05612/) (2024)
- [Preserving Diversity in Supervised Fine-Tuning of Large Language Models](https://mlsanity.com/paper/2408.16673/) (2024)
- [Factuality of Large Language Models: A Survey](https://mlsanity.com/paper/2402.02420/) (2024)
- [Spurious Forgetting in Continual Learning of Language Models](https://mlsanity.com/paper/2501.13453/) (2025)
- [Beyond Factuality: A Comprehensive Evaluation of Large Language Models as Knowledge Generators](https://mlsanity.com/paper/2310.07289/) (2023)
- [How do language models learn facts? Dynamics, curricula and hallucinations](https://mlsanity.com/paper/2503.21676/) (2025)
- [Unveiling the Generalization Power of Fine-Tuned Large Language Models](https://mlsanity.com/paper/2403.09162/) (2024)
- [Understanding Finetuning for Factual Knowledge Extraction](https://mlsanity.com/paper/2406.14785/) (2024)

## Abstract (arXiv)

Supervised fine-tuning (SFT) can degrade factual behavior outside the target domain. This degradation is often described as catastrophic forgetting, yet open-ended factual failures do not necessarily imply that the underlying facts have been erased. In this work, we identify a more specific phenomenon, factual access failure: after domain SFT, models can still recognize or rank the correct answer under constrained evaluation, while failing to produce it in closed-book generation. Through benchmark-level comparisons, same-fact multiple-choice and generation probes, and failure-mode analysis, we show that SFT-induced factual degradation reflects both genuine wrong-answer generations and expression-level failures such as verbosity, formatting mismatch, and exact-match artifacts. To address this problem, we introduce Recall-Anchored Distillation (RAD), a base-anchored self-distillation objective that preserves out-of-distribution generation behavior by aligning the adapted model with the original base model's soft continuation distribution on unlabeled OOD text. RAD requires no gold OOD answers, external judges, or labeled factual data. Across three backbones fine-tuned on MedMCQA, RAD recovers a consistent portion of the lost OOD recall while preserving target-domain adaptation. Compared with replay on the same OOD text, RAD shows that the key preservation signal is the base model's soft distribution rather than additional text exposure alone.