# KNIFE: Distilling Reasoning Knowledge From Free-Text Rationales

arXiv: https://arxiv.org/abs/2212.09721 · HTML: https://mlsanity.com/paper/2212.09721/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning.
- Builds on 22 of its 46 references that are in the map.
- First posted on arXiv 19 Dec 2022.

## Placed in

- LLM reasoning / [Reasoning benchmarks and evaluation](https://mlsanity.com/lr/reasoning-benchmarks-and-evaluation/)
- Knowledge distillation / [Knowledge distillation for LLMs](https://mlsanity.com/distill/knowledge-distillation-for-llms/)

## Builds on (most cited first, 22 of 22)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter](https://mlsanity.com/paper/1910.01108/) (2019)
- [Emergent Abilities of Large Language Models](https://mlsanity.com/paper/2206.07682/) (2022)
- [Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering](https://mlsanity.com/paper/1809.02789/) (2018)
- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [GPT-NeoX-20B: An Open-Source Autoregressive Language Model](https://mlsanity.com/paper/2204.06745/) (2022)
- [Large Language Models Can Self-Improve](https://mlsanity.com/paper/2210.11610/) (2022)
- [Teaching Small Language Models to Reason](https://mlsanity.com/paper/2212.08410/) (2022)
- [Unifying distillation and privileged information](https://mlsanity.com/paper/1511.03643/) (2015)
- [Unsupervised Commonsense Question Answering with Self-Talk](https://mlsanity.com/paper/2004.05483/) (2020)
- [Distilling Reasoning Capabilities into Smaller Language Models](https://mlsanity.com/paper/2212.00193/) (2022)
- [WT5?! Training Text-to-Text Models to Explain their Predictions](https://mlsanity.com/paper/2004.14546/) (2020)
- [Learning by Distilling Context](https://mlsanity.com/paper/2209.15189/) (2022)
- [Explanations from Large Language Models Make Small Reasoners Better](https://mlsanity.com/paper/2210.06726/) (2022)
- [When Can Models Learn From Explanations? A Formal Framework for Understanding the Roles of Explanation Data](https://mlsanity.com/paper/2102.02201/) (2021)
- [PINTO: Faithful Language Reasoning Using Prompt-Generated Rationales](https://mlsanity.com/paper/2211.01562/) (2022)
- [Towards Explainable NLP: A Generative Explanation Framework for Text Classification](https://mlsanity.com/paper/1811.00196/) (2018)
- [REV: Information-Theoretic Evaluation of Free-Text Rationales](https://mlsanity.com/paper/2210.04982/) (2022)
- [Investigating the Benefits of Free-Form Rationales](https://mlsanity.com/paper/2206.11083/) (2022)
- [FRAME: Evaluating Rationale-Label Consistency Metrics for Free-Text Rationales](https://mlsanity.com/paper/2207.00779/) (2022)
- [ER-Test: Evaluating Explanation Regularization Methods for Language Models](https://mlsanity.com/paper/2205.12542/) (2022)

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

- [ELAD: Explanation-Guided Large Language Models Active Distillation](https://mlsanity.com/paper/2402.13098/) (2024)

## Abstract (arXiv)

Language models (LMs) have yielded impressive results on many language reasoning tasks, but their unexpected errors raise doubts about their reasoning abilities. In light of this, there is growing interest in finetuning/prompting LMs with both task instances and their associated free-text rationales (FTRs), which explain the correct reasoning process for predicting the correct task output (i.e., how to be "right for the right reasons"). However, existing finetuning methods fail to improve LM performance, while prompting needs prohibitively large (i.e., >50B) LMs to work well. We propose KNIFE, which shows that reasoning knowledge can be effectively distilled from FTRs into a small (i.e., <1B) LM and improve the LM's performance. First, KNIFE finetunes a teacher LM (given task input and FTR) to predict the task output, transferring reasoning knowledge from the FTRs to the teacher's hidden states. Second, KNIFE finetunes a student LM (given task input only) such that its hidden states are aligned with the teacher's. Thus, the student is endowed with reasoning knowledge but can be used for inference without direct FTR input. On two question-answering datasets, KNIFE outperforms various finetuning and prompting baselines in fully-supervised and low-resource settings. Also, we observe that FTR quality is crucial to KNIFE's performance.