AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

LIREx: Augmenting Language Inference with Relevant Explanations

Xinyan Zhao, V.G.Vinod Vydiswaran

AAAI 2021 · Large language models › Language, safety and interpretability · approach: Faithfulness of LLM explanations · 19 citations (OpenAlex)

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add probability-sampled, rationale-enabled explanations to the P-H inputLIREx on SNLI · SNLI testaccuracy: 91.55{ $\pm$.04} -> 91.87{ $\pm$.03}Improved
Use probability-based sampling rather than greedy highest-probability selection (ablation)LIREx on SNLI · SNLI testaccuracy: 91.73{ $\pm$.03} -> 91.87{ $\pm$.03}Improved
Use explanations alone rather than combining them with the P-H pair (ablation)LIREx on SNLI · SNLI testaccuracy: 91.87 -> 62.21Hurt
Remove explanations from the inference input (ablation)LIREx on SNLI · SNLI testaccuracy: 91.87 -> 91.07Hurt
Use NILE-generated explanations instead of LIREx-generated explanations (ablation)LIREx architecture · SNLI test, random explanation selectionaccuracy: 91.55 -> 85.58Hurt
Randomly select an explanation at evaluation instead of using the best explanation (ablation)LIREx · SNLI testaccuracy: 91.80 -> 91.55Hurt
Use probability-sampled rationale-enabled explanations with P-H inputLIREx transferred from SNLI to MultiNLI · MultiNLI Dev-matchedaccuracy: 80.12 -> 79.85Hurt
Use probability-sampled rationale-enabled explanations with P-H inputLIREx transferred from SNLI to MultiNLI · MultiNLI Dev-mismatchedaccuracy: 79.73 -> 79.79Improved
Remove label information from the explanation-generation prompt (ablation)LIREx explanation generator · SNLIaccuracy: 91.59 -> 91.80Improved
Add three explanations to the P-H input using NILE_all (ablation)NILE on SNLI · SNLI testaccuracy: 91.49 -> 91.12Hurt
Add extra negative samples to NILE_all training (ablation)NILE on SNLI · SNLI testaccuracy: 91.12 -> 90.73Hurt
Use the selected explanation without the P-H pair (ablation)LIREx on SNLI · SNLI testaccuracy: 91.87 -> 89.73Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy