Challenge: Large language models (LLMs) are increasingly used as reasoning engines in autonomous driving, yet their decision-making remains opaque.
Approach: They propose to use gradient-based optimization on continuous embeddings to identify the input shifts required to flip a model’s decision.
Outcome: The proposed method exposes latent biases and provides concrete insights to improve the robustness of LLM-based driving agents.

Similar Papers

Can LLMs Explain Themselves Counterfactually? (2025.emnlp-main)

Copied to clipboard

Challenge: Explanations are an important tool for gaining insights into model behavior, calibrating user trust, and ensuring compliance.
Approach: They propose to use self-explanation to prompt models to explain outputs . they find that LLMs struggle to generate SCEs - their prediction often does not agree with their own counterfactual reasoning.
Outcome: The proposed methods can generate SCEs across families, sizes, temperatures, and datasets.
LLMs Don’t Know Their Own Decision Boundaries: The Unreliability of Self-Generated Counterfactual Explanations (2025.emnlp-main)

Copied to clipboard

Challenge: Existing studies on language models' ability to explain their decisions in natural language have focused on self-generated counterfactual explanations (SCEs).
Approach: They evaluate whether LLMs can generate valid counterfactuals and minimal ones . authors suggest that SCEs are, at best, an ineffective explainability tool .
Outcome: The proposed language models can explain their decisions in natural language, the study finds . the models can produce valid counterfactual explanations, but make small edits that fail to change predictions.
Parallel Universes, Parallel Languages: A Comprehensive Study on LLM-based Multilingual Counterfactual Example Generation (2026.acl-long)

Copied to clipboard

Challenge: Large language models excel at generating English counterfactuals but their effectiveness in generating multilingual counterfacts remains unclear.
Approach: They conduct automatic evaluations on both directly generated and derived counterfactuals in six languages and find that cross-lingual perturbations follow common strategic principles.
Outcome: The proposed models show that translation-based counterfactuals offer higher validity than their directly generated counterparts, but still fall short of matching the quality of the original English counterf actuals.
LLMs for Generating and Evaluating Counterfactuals: A Comprehensive Study (2024.findings-emnlp)

Copied to clipboard

Challenge: Large Language Models (LLMs) have shown remarkable performance in NLP tasks, but their efficacy in generating high-quality CFs remains uncertain.
Approach: They compare LLMs' ability to generate CFs that flip the original label and human CF's.
Outcome: The proposed models generate fluent CFs, but struggle to keep the induced changes minimal.
CATfOOD: Counterfactual Augmented Training for Improving Out-of-Domain Performance and Calibration (2024.eacl-long)

Copied to clipboard

Challenge: Large language models (LLMs) have shown remarkable generalization capabilities, performing well on various tasks such as question answering (QA), complex reasoning, and code generation.
Approach: They propose to augment training data of smaller language models with automatically generated counterfactuals (CF) instances to improve out-of-domain (OOD) performance of SLMs in extractive question answering setup.
Outcome: The proposed approach improves out-of-domain (OOD) performance of small language models in extractive question answering setup.
Dually Self-Improved Counterfactual Data Augmentation Using Large Language Model (2025.acl-long)

Copied to clipboard

Challenge: Existing approaches to generate counterfactual data augmentation are limited due to imbalance and biases in real-world training data.
Approach: They propose a self-improved method for generating high-quality counterfacts using large language models.
Outcome: The proposed method generates high-quality counterfacts on the natural language inference task using lightweight and task-specific LLMs.
Counterfactuals of Counterfactuals: a back-translation-inspired approach to analyse counterfactual editors (2023.findings-acl)

Copied to clipboard

Challenge: Existing explanations for classifiers are counterfactual or contrastive . lack of universal ground truth for counterf actual edits hinders their evaluation .
Approach: They propose a back translation-inspired evaluation methodology that utilises earlier outputs of the explainer as ground truth proxies to investigate the consistency of explainers.
Outcome: The proposed method can provide valuable insights into the behaviour of predictor and explainer models and infer patterns that would otherwise be obscured.
DoCoGen: Domain Counterfactual Generation for Low Resource Domain Adaptation (2022.acl-long)

Copied to clipboard

Challenge: Existing domain adaptation (DA) algorithms are not able to handle out-of-distribution examples due to the costly and labor-intensive data labeling process.
Approach: They propose a controllable generation approach to deal with domain adaptation challenge by generating a domain-counterfactual textual example from an input text.
Outcome: The proposed approach outperforms baselines and improves accuracy of state-of-the-art unsupervised DA algorithm.
SAGE: An Agentic Explainer Framework for Interpreting SAE Features in Language Models (2026.eacl-industry)

Copied to clipboard

Challenge: Large language models (LLMs) have achieved remarkable progress, yet their internal mechanisms remain largely opaque.
Approach: They propose an agent-based framework that recasts feature interpretation from a passive, single-pass generation task into an explanation-driven process.
Outcome: The proposed framework produces explanations with significantly higher generative and predictive accuracy compared to state-of-the-art baselines.
Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs (2026.findings-eacl)

Copied to clipboard

Challenge: Despite efforts to mitigate social bias in large language models, representational harms such as stereotyping continue to exist in both open and closed-source models.
Approach: They propose a method to modify model activations in forward passes by applying steering vectors to a BBQ dataset and comparing their results to bias mitigation methods.
Outcome: The proposed method outperforms 3 other bias mitigation methods on the BBQ dataset and shows the lowest impact on MMLU scores.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations