Papers with model sensitivity
BERT, are you paying attention? Attention regularization with human-annotated rationales (2026.eacl-long)
Copied to clipboard
| Challenge: | Attention regularisation aims to supervise the attention patterns in language models like BERT. |
| Approach: | They compare regularisation on human rationales with random tokens to find that human-annotated rationale is better at reducing model sensitivity to spurious correlations. |
| Outcome: | The proposed regularisation method improves model performance and model robustness, but not with human-annotated rationales. |
Do GUI Grounders Truly Understand UI Elements? (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing grounding models and benchmarks are skewed toward web and mobile environments, neglecting desktop interfaces (especially windows). |
| Approach: | They propose a GUI Grounding Sensitivity Benchmark to assess UI grounding sensitivity to multiple descriptions of the same UI element. |
| Outcome: | The proposed model generates multiple valid instructions per UI element and develops nuanced validation methods to validate them. |
Sensitivity, Performance, Robustness: Deconstructing the Effect of Sociodemographic Prompting (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing studies on sociodemographic prompting have not explored the effectiveness of this technique. |
| Approach: | They propose to use sociodemographic prompting to steer models towards answers that humans with specific sociodemography would give. |
| Outcome: | The proposed technique can improve zero-shot learning by focusing on human sociodemographic profiles. |
Characterizing LLM Abstention Behavior in Science QA with Context Perturbations (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Prior work has investigated the ability of LLMs to abstain from answering context-dependent questions when provided insufficient or inconsistent context is provided. |
| Approach: | They propose to improve abstention when provided insufficient or incorrect context . they probed the ability of LLMs to abstain from answering context-dependent science questions . |
| Outcome: | The proposed models abstain from answering science questions when provided insufficient or incorrect context. |
LIBERTy: A Causal Framework for Benchmarking Concept-Based Explanations of LLMs with Structural Counterfactuals (2026.findings-acl)
Copied to clipboard
| Challenge: | Concept-based explanations quantify how high-level concepts influence model behavior . existing benchmarks rely on costly human-written counterfactuals that serves as imperfect proxy . |
| Approach: | They propose a framework for constructing datasets containing structural counterfactual pairs . they use a structured Causal Model to generate a concept-based explanation . |
| Outcome: | The proposed framework compares concept-based explanations to causal effects estimated from counterfactuals. |