Challenge: Recent advances in reasoning-oriented models have demonstrated impressive capabilities in mathematical reasoning, but their ability to adhere to user directives remains underexplored.
Approach: They propose a benchmark to evaluate instruction-following in mathematical reasoning tasks.
Outcome: The proposed model degrades in instruction adherence when generation length increases, but can partially recover obedience, despite increasing generation length.

Similar Papers

ReasonIF: Large Reasoning Models Fail to Follow Instructions During Reasoning (2026.findings-acl)

Copied to clipboard

Challenge: Prior studies assess instruction adherence in the model’s main responses, but it is also critical for large reasoning models to follow user instructions throughout their reasoning process.
Approach: They propose a systematic benchmark for assessing reasoning instruction following to assess the model's adherence to instructions.
Outcome: The proposed benchmark reduces the risk of undesirable shortcuts, hallucinations, or reward hacking within reasoning traces.
Reasoning Up the Instruction Ladder for Controllable Language Models (2026.findings-acl)

Copied to clipboard

Challenge: Current models struggle to balance competing directives, causing conflicting instructions.
Approach: They propose to reframe instruction hierarchy resolution as a reasoning task . they use a training dataset to enable this capability by transferring general reasoning capabilities to instruction prioritization .
Outcome: The proposed method improves on safety-critical scenarios beyond the training distribution and jailbreaks.
Scaling Behaviors of LLM Reinforcement Learning Post-Training: An Empirical Study in Mathematical Reasoning (2026.acl-long)

Copied to clipboard

Challenge: elucidating scaling laws for large language models (LLMs) during pre-training remains unexplored.
Approach: They characterize how model scale, data, and compute interact during pre-training . they find that large models consistently demonstrate superior compute and data efficiency .
Outcome: The proposed scaling laws offer practical guidance for scaling reasoning capabilities through reinforcement learning post-training.
CodeIF: Benchmarking the Instruction-Following Capabilities of Large Language Models for Code Generation (2025.acl-industry)

Copied to clipboard

Challenge: CodeIF assesses the ability of large language models to adhere to task-oriented instructions in code generation tasks.
Approach: They introduce a benchmark designed to assess LLMs' ability to adhere to task-oriented instructions within diverse code generation scenarios.
Outcome: The proposed benchmark assesses LLMs' ability to adhere to task-oriented instructions in code generation tasks across a wide range of complexity levels and programming domains.
ImpRIF: Stronger Implicit Reasoning Leads to Better Complex Instruction Following (2026.acl-long)

Copied to clipboard

Challenge: Experiments show that enhancing implicit reasoning capabilities can significantly improve complex instruction following in large language models.
Approach: They propose a method to enhance LLMs’ understanding of implicit reasoning instructions by formalizing such instructions as verifiable reasoning graphs and fine-tuning with graph reasoning.
Outcome: The proposed method outperforms existing models on five complex instruction following benchmarks and will be open-sourced in the near future.
PARIF: Pushing the Pareto Frontier of Instruction Following and Reasoning with Curriculum Reinforcement Learning (2026.acl-long)

Copied to clipboard

Challenge: Existing alignment methods struggle to balance general reasoning with instruction-following (IF) this is hindered by dependency on teacher models, reward hacking, and reasoning-answer inconsistencies.
Approach: They propose a two-stage curriculum learning framework based on Reinforcement Learning from Verifiable Rewards to enhance both IF and general reasoning capabilities.
Outcome: The proposed framework outperforms leading models on six representative IF tasks while achieving a 21.25% relative average improvement over the original model.
Working Memory Identifies Reasoning Limits in Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Using large language models, we examine the limitations of their cognitive capabilities and their working memory.
Approach: They examine the limitations of large language models from a scaling perspective . they also assess various prompting strategies, revealing their diverse impacts on LLM performance.
Outcome: The proposed models perform poorly on n-back tasks and on prompting strategies.
Find the Intention of Instruction: Comprehensive Evaluation of Instruction Understanding for Large Language Models (2025.findings-naacl)

Copied to clipboard

Challenge: LLMs are prone to generate responses to instruction-formatted statements in an instinctive manner, rather than comprehending the underlying user intention within the given instructions.
Approach: They propose to use an instruction-following capability benchmark to evaluate LLMs' instruction understanding capability.
Outcome: The proposed benchmark analyzes the instruction understanding capability of large language models with four instruction candidates and a single candidate.
The SIFo Benchmark: Investigating the Sequential Instruction Following Ability of Large Language Models (2024.findings-emnlp)

Copied to clipboard

Challenge: Current evaluation resources for instruction following focus on single task instructions, but the instruction sequences in these benchmarks often lack coherence.
Approach: They propose to evaluate models’ abilities to follow multiple instructions through sequential instruction following tasks using four tasks to assess different aspects of sequential instruction followed.
Outcome: The proposed benchmark outperforms open-source and closed-source models on four tasks assessing different aspects of sequential instruction following.
Dancing in Chains: Reconciling Instruction Following and Faithfulness in Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Modern language models fail to follow human instructions while being faithful . a trade-off exists between instruction following and faithfulness when training LMs .
Approach: They propose a method that relies on Reject-sampling by Self-instruct with Continued Fine-tuning to train LMs to follow human instructions while being faithful.
Outcome: The proposed method outperforms vanilla MTL with high-quality data, but with significantly smaller data.

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