Challenge: Current evaluations of large language models rely on a single instruction template, overlooking models’ sensitivity to instruction style.
Approach: They propose a multi-dimensional framework quantifying how instruction formulation affects model responses by transforming benchmark problems into multiple instruction styles.
Outcome: The proposed framework reveals that instruction style can shift accuracy by 16.7% points.

Similar Papers

What Did I Do Wrong? Quantifying LLMs’ Sensitivity and Consistency to Prompt Engineering (2025.naacl-long)

Copied to clipboard

Challenge: Large Language Models (LLMs) have significantly improved productivity in a number of routine tasks.
Approach: They propose two metrics for classification tasks, namely *sensitivity* and *consistency*, which are complementary to task performance.
Outcome: The proposed metrics are complementary to task performance and can be used to guide prompt engineering and obtain LLMs that balance robustness and performance.
SCAR: Data Selection via Style Consistency-Aware Response Ranking for Efficient Instruction-Tuning of Large Language Models (2025.acl-long)

Copied to clipboard

Challenge: Recent studies show that manually ensuring a consistent response style and maintaining high data quality can significantly improve the performance of fine-tuned Large Language Models (LLMs).
Approach: They introduce a style-aware response ranking system that prioritizes instruction-response pairs based on their stylistic consistency.
Outcome: The proposed model matches or surpasses models trained on the entire dataset in coding and open-ended question-answering benchmarks.
Revisiting the Reliability of Language Models in Instruction-Following (2026.acl-long)

Copied to clipboard

Challenge: Several benchmarks have been proposed to measure instruction-following accuracy, but these scores do not translate to reliable services in real-world use.
Approach: They propose a new metric reliable@k and develop an automated pipeline to generate cousin prompts.
Outcome: The proposed model can be instantiated with cousin prompts and generates high-quality cousin prompt data.
MAIN: Mutual Alignment Is Necessary for instruction tuning (2025.emnlp-main)

Copied to clipboard

Challenge: Instruction tuning has enabled large language models to achieve remarkable performance, yet its success heavily depends on the availability of high-quality instruction-response pairs.
Approach: They propose a mutual alignment framework which enforces coherence between instructions and responses through mutual constraints.
Outcome: The proposed framework generalizes well across model architectures and sizes, achieving state-of-the-art performance on LLaMA, Mistral, and Qwen models across diverse benchmarks.
SCORE: Systematic COnsistency and Robustness Evaluation for Large Language Models (2025.naacl-industry)

Copied to clipboard

Challenge: Typical evaluations of Large Language Models (LLMs) report a single accuracy metric per dataset, often derived from an optimized setup.
Approach: They propose a framework for non-adversarial evaluation of large language models that evaluates models by repeatedly testing them on the same benchmarks in various setups.
Outcome: The proposed framework evaluates models by repeatedly testing them on the same benchmarks in various setups to give a realistic estimate of their accuracy and consistency.
How Reliable Are Automatic Evaluation Methods for Instruction-Tuned LLMs? (2024.findings-emnlp)

Copied to clipboard

Challenge: Existing work on instruction-tuned Large Language Models has used automatic methods based on text overlap and LLM judgments as cost-effective alternatives to human evaluation.
Approach: They perform a meta-evaluation of automatic methods and assess their reliability across a broad range of tasks.
Outcome: The proposed method is unreliable in free-form generation tasks and cross-lingual scenarios.
When Instructions Multiply: Measuring and Estimating LLM Capabilities of Multiple Instructions Following (2025.findings-emnlp)

Copied to clipboard

Challenge: a large number of languages are increasingly used to evaluate their ability to follow multiple instructions simultaneously.
Approach: They propose two benchmarks to evaluate LLMs' ability to follow multiple instructions simultaneously . they use many instruction-following eval and style-aware Mostly Basic programming problems .
Outcome: The proposed models predict performance on unseen instruction combinations and not used during training with 10% error.
Influences on LLM Calibration: A Study of Response Agreement, Loss Functions, and Prompt Styles (2025.acl-long)

Copied to clipboard

Challenge: Existing studies neglect to measure the generalization of their methods to other prompt styles and different sizes of LLMs.
Approach: They propose a framework that trains an auxiliary model for confidence estimation that aggregates responses from multiple LLMs to capture inter-model agreement.
Outcome: The proposed framework integrates response agreement and focal loss with binary cross-entropy to improve calibration from baselines.
Measuring What Matters: Evaluating Ensemble LLMs with Label Refinement in Inductive Coding (2025.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) are prone to inconsistencies and individual biases, limiting their reliability.
Approach: They propose a framework that combines ensemble methods with code refinement methodology to address these challenges.
Outcome: The proposed framework outperforms large language models and LLMs with a low-rank averaging and a moderator-based mechanism to simulate human consensus.
Can LLMs Compress (and Decompress)? Evaluating Code Understanding and Execution via Invertibility (2026.findings-acl)

Copied to clipboard

Challenge: a recent development of code-LLMs has demonstrated remarkable performance across various software engineering applications.
Approach: They propose a round-trip code execution reasoning task to test round- trip consistency . they use zero-shot prompting, supervised fine-tuning on execution traces and self-reflection mechanisms to evaluate models .
Outcome: The proposed benchmarks show that LLMs struggle with round-trip consistency . the benchmarks lack the internal coherence required for trustworthy code reasoning .

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