Increasing Probability Mass on Answer Choices Does Not Always Improve Accuracy (2023.emnlp-main)
Copied to clipboard
| Challenge: | Pretrained language models (LMs) are used to discriminate on multiple-choice tasks that place probability mass on vocabulary tokens that aren’t among the given answer choices. |
| Approach: | They propose a mathematical formalism for SFC which allows us to quantify and bound its impact for the first time. |
| Outcome: | The proposed method eliminates the impact of SFC in the majority of instances. |
Similar Papers
S2LPP: Small-to-Large Prompt Prediction across LLMs (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a small model can be used to select effective prompt templates for a larger model. |
| Approach: | They propose a method to use a smaller model to select effective prompt templates for a larger model. |
| Outcome: | The proposed method significantly reduces the cost of prompt engineering while matching performance with optimal prompts among candidates. |
Large Language Models Might Not Care What You Are Saying: Prompt Format Beats Descriptions (2025.findings-emnlp)
Copied to clipboard
| Challenge: | In-context learning has improved performance of large language models, but descriptive instructions are still under-explored. |
| Approach: | They propose an ensemble prompt framework to describe selection criteria of multiple in-context examples. preliminary experiments on machine translation confirm that this framework boosts ICL performance. |
| Outcome: | The proposed framework improves on commonsense, math, logical reasoning and hallucination tasks with three LLMs. |
ProFit: Leveraging High-Value Signals in SFT via Probability-Guided Token Selection (2026.findings-acl)
Copied to clipboard
| Challenge: | Traditional fine-tuning ignores one-to-many nature of language, leading to overfitting . authors propose a method to fine- tune LLMs by leveraging tokens. |
| Approach: | They propose a method to fine-tune Large Language Models by leveraging tokens to mask low-probability tokens. |
| Outcome: | The proposed method outperforms baselines on general reasoning and mathematical benchmarks. |
“My Answer is C”: First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models (2024.findings-acl)
Copied to clipboard
Xinpeng Wang, Bolei Ma, Chengzhi Hu, Leon Weber-Genzel, Paul Röttger, Frauke Kreuter, Dirk Hovy, Barbara Plank
| Challenge: | Multiple choice questions are one of the most popular evaluation formats for understanding the capabilities of autoregressive large language models (LLMs). |
| Approach: | They evaluated how aligned first-token evaluation is with the text output along several dimensions, namely final option choice, refusal rate, choice distribution and robustness under prompt perturbation. |
| Outcome: | The proposed evaluation methods are misaligned on all dimensions, reaching mismatch rates over 60%. |
Precise Task Formalization Matters in Winograd Schema Evaluations (2020.emnlp-main)
Copied to clipboard
| Challenge: | Recent results report a surge in performance to nearhuman levels on the Winograd Schema Challenge (WSC) however, variations in task formulation across papers and evaluations makes it hard to understand the true degree of recent progress. |
| Approach: | They propose to use a model with multiple choice to frame the task as multiple choice and reuse a pretrained language modeling head to mitigate the model's extreme sensitivity to hyperparameters. |
| Outcome: | The proposed frameworks improve the model's reasoning ability by framing the task as multiple choice and reuse of a pretrained language modeling head. |
Characterizing Mechanisms for Factual Recall in Language Models (2023.emnlp-main)
Copied to clipboard
| Challenge: | Language Models often integrate facts they memorized with new information that appears in a given context, causing competition within the model. |
| Approach: | They investigate distributional and mechanistic determinants of LM behavior in a dataset that queries for knowledge of world capitals . they use head attribution to identify individual attention heads that either promote the memorized answer or the in-context answer in the logits . |
| Outcome: | The proposed method can increase the rate of generating the in-context answer to 88% of the time by scaling up or down the value vector of individual attention heads at runtime. |
Rethinking Pragmatics in Large Language Models: Towards Open-Ended Evaluation and Preference Tuning (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods to assess social-pragmatic inference in large language models are inadequacy, and preferential tuning is the best approach. |
| Approach: | They propose to use free-form models' responses as a measure to assess social-pragmatic reasoning and advocate for preference optimization over supervised finetuning (SFT). |
| Outcome: | The proposed model outperforms supervised finetuning (SFT) and offers a near-free launch in pragmatic abilities without compromising general capabilities. |
Tokenizer Choice For LLM Training: Negligible or Crucial? (2024.findings-naacl)
Copied to clipboard
Mehdi Ali, Michael Fromm, Klaudia Thellmann, Richard Rutmann, Max Lübbering, Johannes Leveling, Katrin Klug, Jan Ebert, Niclas Doll, Jasper Buschhoff, Charvi Jain, Alexander Weber, Lena Jurkschat, Hammam Abdelwahab, Chelsea John, Pedro Ortiz Suarez, Malte Ostendorff, Samuel Weinbach, Rafet Sifa, Stefan Kesselheim, Nicolas Flores-Herr
| Challenge: | Recent success of large language models has been driven by curating the training dataset composition, scaling of model architectures and advancements in pretraining objectives, leaving tokenizer influence as a blind spot. |
| Approach: | They conduct a comprehensive study on the influence of tokenizer choice on LLM downstream performance by training 24 mono- and multilingual LLMs at a 2.6B parameter scale. |
| Outcome: | The proposed model can significantly impact the model's downstream performance and training costs. |
Not-Just-Scaling Laws: Towards a Better Understanding of the Downstream Impact of Language Model Design Decisions (2025.emnlp-main)
Copied to clipboard
Emmy Liu, Amanda Bertsch, Lintang Sutawika, Lindia Tjuatja, Patrick Fernandes, Lara Marinov, Michael Chen, Shreya Singhal, Carolin Lawrence, Aditi Raghunathan, Kiril Gashteovski, Graham Neubig
| Challenge: | Language model performance is largely dependent on pretraining decisions, but scaling laws based on only these two aspects do not always explain downstream task performance. |
| Approach: | They meta-analyze 92 open-source pretrained models to quantify their impact on performance. |
| Outcome: | The framework lays a foundation for more systematic investigation of how model development choices shape final capabilities. |
Selecting Better Samples from Pre-trained LLMs: A Case Study on Question Generation (2023.findings-acl)
Copied to clipboard
Xingdi Yuan, Tong Wang, Yen-Hsiang Wang, Emery Fine, Rania Abdelghani, Hélène Sauzéon, Pierre-Yves Oudeyer
| Challenge: | Large Language Models (LLMs) have demonstrated impressive prowess in natural language generation. |
| Approach: | They propose a method to select high-quality questions from LLM-generated candidates using round-trip and prompt-based scoring. |
| Outcome: | The proposed approach can select high-quality questions from a set of LLM-generated candidates without modification of the underlying model nor rely on human annotations. |