Though Language Models Err While They Strive: Conformal Prediction for Self-Correcting Scientific Generation
Summary
Introduces Scientific Feasibility Control (SFC), a conformal prediction framework that provides statistical guarantees for scientific reasoning validity in LLMs, achieving 50.1% on PhyX physics reasoning, outperforming DeepSeek-R1 and GPT-4 while reducing scientific violations by 73%.
View Cached Full Text
Cached at: 07/21/26, 06:43 AM
# Though Language Models Err While They Strive: Conformal Prediction for Self-Correcting Scientific Generation Source: [https://arxiv.org/abs/2607.16704](https://arxiv.org/abs/2607.16704) [View PDF](https://arxiv.org/pdf/2607.16704) > Abstract:Large language models frequently violate fundamental scientific principles when generating technical content, undermining their reliability in scientific applications\. We introduce Scientific Feasibility Control SFC, a graph\-structured conformal prediction framework that provides statistical guarantees for scientific reasoning validity through progressive absolute\-coherent\-factuality validation\. Our approach decomposes scientific reasoning into atomic absolute\-coherent\-factuality units requiring both individual correctness against physical laws and logical substantiation from preceding context, addressing the cascade effect where early scientific errors contaminate subsequent reasoning steps\. Unlike independence\-based methods that treat claims in isolation, SFC models logical dependencies as approximate deducibility graphs and operates through real\-time validation with dynamic branching when scientific violations are detected, the system branches to alternative generation paths using verified context as foundation\. We demonstrate SFC across established scientific reasoning benchmarks including PhyX multimodal physics, MATH, ScienceQA, and ARC Challenge, achieving 50\.1 percent accuracy on PhyX physics reasoning, substantially outperforming recent reasoning models including DeepSeek\-R1 49\.8 percent and GPT\-4 45\.8 percent while providing 91\.7 percent scientific validity with formal conformal coverage guarantees at alpha equals 0\.10 confidence level and reducing scientific law violations by 73 percent across multiple model architectures\. ## Submission history From: Hao Zhang \[[view email](https://arxiv.org/show-email/7a9df715/2607.16704)\] **\[v1\]**Sat, 18 Jul 2026 08:38:19 UTC \(389 KB\)
Similar Articles
Experiments or Outcomes? Probing Scientific Feasibility in Large Language Models
UMBC researchers show LLMs judge scientific claim feasibility better when given outcome data than experiment descriptions, and that incomplete experimental context can hurt accuracy.
Assessing Language Models for Scientific Applications: Inertial Confinement Fusion as a Use Case
Researchers from Los Alamos National Laboratory and NVIDIA present an expert-curated benchmark of 250 questions on inertial confinement fusion to evaluate LLMs' scientific fluency, physical reasoning, and quantitative analysis. They compare frontier open/closed models with Prospero, a domain-specialized model, finding that specialization improves scores but critical physics errors persist, and recommend assessing existing capabilities and retrieval benefits before investing in further specialization.
Safety as a Constraint: Fine-Tuning a LLM Recommender to Explain Itself
This paper fine-tunes a large language model for a video streaming service to generate personalized, faithful, and harmless recommendation explanations, improving pass rates without compromising original performance.
When Correct Beliefs Collapse: Epistemic Resilience of LLMs under Clinical Pressure
This paper investigates how large language models maintain correct beliefs under adversarial pressure in clinical settings, proposing R-FT fine-tuning to improve epistemic resilience while balancing corrigibility, and demonstrating significant robustness gains on medical benchmarks.
Beyond Surface Statistics: Robust Conformal Prediction for LLMs via Internal Representations
This paper proposes a conformal prediction framework for LLMs that leverages internal representations rather than output-level statistics, introducing Layer-Wise Information (LI) scores as nonconformity measures to improve validity-efficiency trade-offs under distribution shift. The method demonstrates stronger robustness to calibration-deployment mismatch compared to text-level baselines across QA benchmarks.