WildIFEval: New Dataset Benchmarks Large Language Models on Complex, Multi-Constraint Instructions
Researchers have introduced WildIFEval, a dataset of 7,000 real user instructions with multiple constraints, to evaluate how well large language models follow complex instructions. The dataset categorizes constraints into eight classes and reveals significant performance gaps between small and large models. The work addresses a critical limitation in current LLM evaluation by testing instruction-following under realistic, multi-constraint conditions.
WildIFEval is a large-scale benchmark dataset containing 7,000 real user instructions that incorporate diverse, multi-constraint conditions extracted from natural language interactions. Unlike previous instruction-following datasets, WildIFEval spans a broad lexical and topical spectrum, with constraints organized into eight high-level categories to capture their distribution in real-world scenarios. Extensive experiments benchmarking leading LLMs reveal that all tested models have substantial room for improvement on such tasks, with clear performance differentiation between small and large models. The analysis examines how the number and type of constraints affect model performance, uncovering patterns in how models handle constraint-following behavior. The researchers are releasing the dataset to support further research on instruction-following capabilities under complex, realistic conditions.
What different sources said
- arXiv cs.AICenter
WildIFEval: Instruction Following in the Wild
Related
Topology-Aware Thermodynamics Improves DNA Probe Specificity Design
Researchers developed a new framework for designing DNA probes that accounts for the spatial organization of matched sequences, not just overall thermodynamic stability. Traditional methods rely on scalar measures like melting temperature and free energy, which miss how mismatches are distributed along the probe. The approach could improve diagnostic accuracy in applications like HPV detection and gene expression profiling.
Study Identifies Optimal Thermal Dose for Combining Focused Ultrasound with Immunotherapy in Tumors
Researchers used multimodal PET imaging to identify an optimal thermal dose range for focused ultrasound ablation that destroys tumor tissue while preserving conditions for immunotherapy delivery. The study found that excessive heating collapses blood vessels needed for antibody access, while insufficient heating fails to adequately reduce tumor burden. The findings could guide clinical design of combination treatments pairing thermal ablation with immunotherapies.
Plant MSH1 Protein Functions as Mismatch-Directed Nuclease for Organelle Genome Maintenance
Researchers have identified the precise mechanism by which the AtMSH1 protein in Arabidopsis plants recognizes and cleaves DNA mismatches and lesions, preventing mutations in organellar genomes. The protein combines a DNA mismatch recognition module with a nuclease domain that makes staggered cuts at specific positions relative to DNA damage. This discovery explains how plants maintain unusually low mutation rates in their mitochondrial and chloroplast DNA compared to other eukaryotes.