Aligning What LLMs Do and Say: Towards Self-Consistent Explanations (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are often prompted to produce natural language explanations, but the features driving the answer are often different from those emphasized in their explanations. |
| Approach: | They propose a large-scale benchmark linking model decisions with diverse explanations and attribution vectors across datasets, methods, and model families to address this gap. |
| Outcome: | The proposed model generates an answer where the word NLP in the prompt has high feature importance. |
Similar Papers
Are self-explanations from Large Language Models faithful? (2024.findings-acl)
Copied to clipboard
| Challenge: | Instruction-tuned Large Language Models excel at many tasks and will explain their reasoning, so-called self-explanations. |
| Approach: | They propose to employ self-consistency checks to measure faithfulness to LLMs to determine if they are model-dependent and if their reasoning is convincing and wrong. |
| Outcome: | The proposed measures show that self-explanations are explanation, model, and task-dependent and should not be trusted in general. |
InfoPO: On Mutual Information Maximization for Large Language Model Alignment (2025.naacl-long)
Copied to clipboard
Teng Xiao, Zhen Ge, Sujay Sanghavi, Tian Wang, Julian Katz-Samuels, Marc Versage, Qingjun Cui, Trishul Chilimbi
| Challenge: | Recent studies have shown that direct preference optimization and its variants can be useful for fine-tuning large language models with human preferences data. |
| Approach: | They propose a preference fine-tuning algorithm that effectively and efficiently aligns large language models using preference data. |
| Outcome: | Extensive experiments show that the proposed algorithm outperforms established baselines on reasoning tasks. |
On Measuring Faithfulness or Self-consistency of Natural Language Explanations (2024.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) can explain their predictions through post-hoc or Chain-of-Thought explanations, but they can also make unfaithful explanations that hide their sensitivity to biasing inputs. |
| Approach: | They propose to use a model-based consistency test to judge the faithfulness of post-hoc or Chain-of-Thought explanations rather than model-specific faithfulness tests. |
| Outcome: | The proposed tests do not measure faithfulness to model explainability but rather their self-consistency at output level. |
From Distributional to Overton Pluralism: Investigating Large Language Model Alignment (2025.naacl-long)
Copied to clipboard
| Challenge: | a large language model's (LLM) output distribution is changed by an alignment process . a recent study shows that aligned models surface information that cannot be recovered from base models without fine-tuning. |
| Approach: | They analyze two aspects of the alignment process that change output distributions . they find alignment suppresses irrelevant and unhelpful content . |
| Outcome: | The proposed model can be imitated without fine-tuning by using in-context examples and lower-resolution semantic hints about response content. |
Faithful Serum: Mitigating the Faithfulness Gap in Textual Explanations of LLM Decisions via Attribution Guidance (2026.acl-long)
Copied to clipboard
| Challenge: | Prior work has focused on generating convincing rationales that appear to be subjectively faithful, but it remains unclear whether these explanations are epistemic faithful. |
| Approach: | They propose a method that enhances epistemic faithfulness by guiding explanation generation through attention-level interventions, informed by token-level heatmaps. |
| Outcome: | The proposed method significantly improves epistemic faithfulness across multiple models, benchmarks, and prompts. |
Insights into Alignment: Evaluating DPO and its Variants Across Multiple Tasks (2025.acl-srw)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) excel in math reasoning problemsolving, text generation, summarization, creative writing, among other tasks. |
| Approach: | They evaluate Direct Preference Optimization and its variants for aligning Large Language Models with human preferences. |
| Outcome: | The proposed alignment methods achieve near-optimal performance even with smaller subsets of training data. |
Probability-Consistent Preference Optimization for Enhanced LLM Reasoning (2025.findings-acl)
Copied to clipboard
Yunqiao Yang, Houxing Ren, Zimu Lu, Ke Wang, Weikang Shi, Aojun Zhou, Junting Pan, Mingjie Zhan, Hongsheng Li
| Challenge: | Recent advances in preference optimization have demonstrated significant potential for improving mathematical reasoning capabilities in large language models. |
| Approach: | They propose a framework that establishes two quantitative metrics for preference selection: surface-level answer correctness and intrinsic token-level probability consistency. |
| Outcome: | The proposed framework outperforms existing outcome-only criterion approaches across a diverse range of LLMs and benchmarks. |
Rethinking the Evaluation of Alignment Methods: Insights into Diversity, Generalisation, and Safety (2026.eacl-srw)
Copied to clipboard
Denis Janiak, Julia Moska, Dawid Motyka, Karolina Seweryn, Paweł Walkowiak, Bartosz Żuk, Arkadiusz Janz
| Challenge: | Existing studies focus on individual techniques or specific dimensions, lacking a holistic assessment of the inherent trade-offs. |
| Approach: | They propose a framework that compares LLM alignment methods across five axes . they use a validated LLM-as-judge prompt to compare the results . |
| Outcome: | The proposed framework compares LLM alignment methods across factuality, safety, conciseness, proactivity, diversity and safety axes . it provides insights into trade-offs of common alignment methods, guiding the development of more balanced and reliable LLMs. |
Bridging Internal Consistency and External Alignment: A Causal and Dynamic Interpretability Framework for LLM Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing interpretability methods focus on internal and external aspects of the model . existing explanations often focus on surface correlations or static dependencies . |
| Approach: | They propose a causal and dynamic interpretability framework for Large Language Models . they characterize backdoor-adjusted causal effects of generated prefix and prompt . |
| Outcome: | The proposed framework provides a unified causal view of internal consistency and external alignment in LLM generation dynamics. |
Uncovering Factor-Level Preference to Improve Human-Model Alignment (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models exhibit tendencies that diverge from human preferences, such as favoring certain writing styles or producing overly verbose outputs. |
| Approach: | They propose a framework to uncover and measure factor-level preference alignment of humans and large language models (LLMs) |
| Outcome: | The proposed framework uncovers and measures factor-level preference alignment of humans and large language models. |