SciR: New Benchmark Tests LLMs on Scientific Reasoning with Controllable Difficulty
Researchers introduced SciR, a new benchmark designed to evaluate large language models on three forms of scientific reasoning: deduction, induction, and causal abduction. The benchmark uses formal logical structures rendered into realistic scientific documents, allowing independent control over information extraction difficulty and inference difficulty. This addresses a gap in existing benchmarks, which either rely on costly human annotations or use synthetic logic problems that don't resemble real scientific literature.
SciR combines multi-paradigm reasoning tasks with controllable scientific document rendering to create a more rigorous evaluation framework for LLMs. The benchmark generates tasks from formal objects—deduction trees, inductive rule hypotheses, and causal graphs—ensuring verifiable ground-truth answers, then renders them into multi-document scientific discourse using domain-tuned genres. Testing six models revealed that both difficulty axes (information extraction and inference) negatively impact performance, with compounding effects. Notably, the rendering challenges affected even neurosymbolic pipelines that delegate inference to verified solvers. The analysis produced per-model profiles showing that reasoning-specialized models like deepseek-r1 outperform standard instruction-tuned models on inference tasks, though both struggle with extraction difficulty.
What's missing
The paper does not discuss potential limitations of the benchmark design itself, such as whether the three paradigmatic forms of inference fully capture the breadth of scientific reasoning, or how the benchmark might generalize to scientific domains not represented in the test set.
What different sources said
- arXiv cs.AICenter
SciR: A Controllable Benchmark for Scientific Reasoning in LLMs
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.