# Inverse IFEval: Can LLMs Unlearn Stubborn Training Conventions to Follow Real Instructions?

arXiv: https://arxiv.org/abs/2509.04292 · HTML: https://mlsanity.com/paper/2509.04292/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Builds on 30 of its 33 references that are in the map.
- First posted on arXiv 4 Sep 2025.

## Builds on (most cited first, 30 of 30)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [Instruction-Following Evaluation for Large Language Models](https://mlsanity.com/paper/2311.07911/) (2023)
- [Instruction Tuning with GPT-4](https://mlsanity.com/paper/2304.03277/) (2023)
- [From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline](https://mlsanity.com/paper/2406.11939/) (2024)
- [GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models](https://mlsanity.com/paper/2508.06471/) (2025)
- [Measuring short-form factuality in large language models](https://mlsanity.com/paper/2411.04368/) (2024)
- [ToolQA: A Dataset for LLM Question Answering with External Tools](https://mlsanity.com/paper/2306.13304/) (2023)
- [Benchmarking Complex Instruction-Following with Multiple Constraints Composition](https://mlsanity.com/paper/2407.03978/) (2024)
- [Beyond Functional Correctness: Exploring Hallucinations in LLM-Generated Code](https://mlsanity.com/paper/2404.00971/) (2024)
- [GLoRe: When, Where, and How to Improve LLM Reasoning via Global and Local Refinements](https://mlsanity.com/paper/2402.10963/) (2024)
- [VisIT-Bench: A Benchmark for Vision-Language Instruction Following Inspired by Real-World Use](https://mlsanity.com/paper/2308.06595/) (2023)
- [A Comprehensive Survey on Process-Oriented Automatic Text Summarization with Exploration of LLM-Based Methods](https://mlsanity.com/paper/2403.02901/) (2024)
- [Can ChatGPT Defend its Belief in Truth? Evaluating LLM Reasoning via Debate](https://mlsanity.com/paper/2305.13160/) (2023)
- [AdaCoT: Pareto-Optimal Adaptive Chain-of-Thought Triggering via Reinforcement Learning](https://mlsanity.com/paper/2505.11896/) (2025)
- [Can ChatGPT Replace Traditional KBQA Models? An In-depth Analysis of the Question Answering Performance of the GPT LLM Family](https://mlsanity.com/paper/2303.07992/) (2023)
- [CIEM: Contrastive Instruction Evaluation Method for Better Instruction Tuning](https://mlsanity.com/paper/2309.02301/) (2023)
- [Instruction Tuning With Loss Over Instructions](https://mlsanity.com/paper/2405.14394/) (2024)
- [INSTRUCTIR: A Benchmark for Instruction Following of Information Retrieval Models](https://mlsanity.com/paper/2402.14334/) (2024)
- [M-IFEval: Multilingual Instruction-Following Evaluation](https://mlsanity.com/paper/2502.04688/) (2025)
- [FollowEval: A Multi-Dimensional Benchmark for Assessing the Instruction-Following Capability of Large Language Models](https://mlsanity.com/paper/2311.09829/) (2023)
- [CIF-Bench: A Chinese Instruction-Following Benchmark for Evaluating the Generalizability of Large Language Models](https://mlsanity.com/paper/2402.13109/) (2024)
- [IFEval-Audio: Benchmarking Instruction-Following Capability in Audio-based Large Language Models](https://mlsanity.com/paper/2505.16774/) (2025)

## Built on by (most cited first, 10 of 10)

- [Seed1.8 Model Card: Towards Generalized Real-World Agency](https://mlsanity.com/paper/2603.20633/) (2026)
- [KAT-Coder Technical Report](https://mlsanity.com/paper/2510.18779/) (2025)
- [Conflict-Aware Fusion: Mitigating Logic Inertia in Large Language Models via Structured Cognitive Priors](https://mlsanity.com/paper/2512.06393/) (2025)
- [MM-CondChain: A Programmatically Verified Benchmark for Visually Grounded Deep Compositional Reasoning](https://mlsanity.com/paper/2603.12266/) (2026)
- [OmniCap-IF: Benchmarking and Improving Instruction Following Abilities for Omni-Video Captioning](https://mlsanity.com/paper/2606.08572/) (2026)
- [PlanningBench: Generating Scalable and Verifiable Planning Data for Evaluating and Training Large Language Models](https://mlsanity.com/paper/2605.20873/) (2026)
- [Overconfident and Blind to Details: Fixing Prompt Insensitivity with Abductive Preference Learning](https://mlsanity.com/paper/2510.09887/) (2025)
- [EvolveScaler: Synthesizing Information-Evolution Contexts via Executable State Machines and Natural-Language Rendering](https://mlsanity.com/paper/2609.08435/) (2026)
- [Learning to Learn from Context: Synthetic Training from Perturbed Public Documents](https://mlsanity.com/paper/2609.33642/) (2026)
- [Learned Reporting Preferences in RLVR Can Conflict with the Current Request](https://mlsanity.com/paper/2609.36587/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) achieve strong performance on diverse tasks but often exhibit cognitive inertia, struggling to follow instructions that conflict with the standardized patterns learned during supervised fine-tuning (SFT). To evaluate this limitation, we propose Inverse IFEval, a benchmark that measures models Counter-intuitive Abilitytheir capacity to override training-induced biases and comply with adversarial instructions. Inverse IFEval introduces eight types of such challenges, including Question Correction, Intentional Textual Flaws, Code without Comments, and Counterfactual Answering. Using a human-in-the-loop pipeline, we construct a dataset of 1012 high-quality Chinese and English questions across 23 domains, evaluated under an optimized LLM-as-a-Judge framework. Experiments on existing leading LLMs demonstrate the necessity of our proposed Inverse IFEval benchmark. Our findings emphasize that future alignment efforts should not only pursue fluency and factual correctness but also account for adaptability under unconventional contexts. We hope that Inverse IFEval serves as both a diagnostic tool and a foundation for developing methods that mitigate cognitive inertia, reduce overfitting to narrow patterns, and ultimately enhance the instruction-following reliability of LLMs in diverse and unpredictable real-world scenarios.