Neuro-formal verification is introduced, which harnesses that automation for developers of mainstream programming languages and returns a Dafny proof of correctness or of a bug on 57% of the entries at 92% precision, and a CBMC counterexample for 63% of the buggy programs at 90% precision.
Abstract
Formal verification offers the strongest assurance available for software, and verification-aware languages have made its automation real. Yet the benefits reach few mainstream developers, most of whose languages have no verification support. Besides, specifying properties and modeling the environment require expertise in formal methods. Proof is therefore reserved for a few celebrated artifacts, while the production code that ships is attested only through review and testing. We introduce neuro-formal verification (NFV), which harnesses that automation for developers of mainstream programming languages: an AI coding agent translates, an established verifier decides, and a question posed in a mainstream language is answered push-button, at empirical accuracy rather than soundness, with a machine-checked proof. Results on a dataset of correct and incorrect solutions to Python programming problems are encouraging compared to an {llm-as-judge} baseline: NFV returns a Dafny proof of correctness or of a bug on 57% of the entries at 92% precision, and a CBMC counterexample for 63% of the buggy programs at 90% precision.
MAGE explains how externalized knowledge, bounded action, independent evaluation, and retained human authority can compose into a governed engineering environment, and proposes tests of when that environment turns commodity intelligence into durable engineering progress.
James C. Davis, Kelechi G. Kalu, Huiyun Peng et al.· 1 citation
LLMs are increasingly used for code generation, yet they frequently hallucinate non-existent software packages, creating exploitable entry points into the software supply chain. We make four contributions to this problem. First, we show that prior evaluation methodologies systematically inflate hallucination rates by misclassifying standard-library modules as hallucinations in some languages. For Python, the overestimation reaches 9.4 percentage points. Second, we evaluate seven inference-time defenses for mitigating package hallucinations, including five guided decoding strategies (Greedy, Contrastive, DoLa, Nudging, and Active Layer-Contrastive Decoding), an iterative self-refinement approach (Self-Refine), and a Retrieval-Augmented Generation (RAG)-based defense.. Across eight models spanning five families and four programming languages (Python, JavaScript, Ruby, Rust), RAG reduces the package hallucination rate (PHR) in 18 of 32 model--language configurations. Third, we introduce Package Utility (PU) to assess whether defenses preserve valid and task-relevant recommendations. Among strategies evaluated, Greedy decoding provides the strongest average mitigation--utility trade-off. Fourth, we stress-test all strategies under adversarial prompts seeded with fabricated package names and find that PHR surges by up to 45 percentage points relative to standard prompts, with Ruby consistently the most vulnerable language (80.9--95.2\%). Under adversarial conditions, RAG and Self-Refine outperform all decoding-only strategies, indicating that robust defense requires either external grounding or iterative self-verification when prompts are actively hostile. Our results recast package hallucination as both a measurement problem and a decoding-time control problem, and they demonstrate that the choice of defense must be matched to the threat model and recommendation utility.
Albérick Euraste Djiré, Iyiola E. Olatunji, Melissa Tessa et al.· 1 citation
Post-training alignment is often shallow, eroding under fine-tuning. It remains untested as to whether constitutional midtraining interventions can produce durable alignment when cleanly isolated from post-training. We build a 394M-token constitutional corpus from Anthropic's Constitution and apply constitutional midtraining at 120B scale, where principled, values-based content is inserted into midtraining. A 2x2 design (curriculum ordering x deliberative reasoning) was used to produce four constitutionally midtrained conditions, plus a control, which were evaluated on self-generated and established benchmarks including alignment under pressure, value conflict resolution, blackmail, and emergent misalignment. All models were evaluated across three stages: post-midtraining, post-SFT, and post-benign fine-tuning. Constitutionally midtrained models outperformed the control on alignment generalization and durability, notably on blackmail: SFT instilled a blackmail propensity in all models, but constitutional midtraining blunted it, with the advantage surviving benign fine-tuning (-17.5pp). This durability did not extend to settings that required active resistance to in-context pressure or conflict, where the advantage attenuates after SFT. The presence of constitutional content at midtraining also mattered more than its structure, and constitutional midtraining incurred no capability cost, on average, at any stage (MMLU, ARC-Easy, piqa, GSM8K). A modest amount of constitutional content at midtraining could therefore yield broad, persistent alignment gains, offering a cheap, complementary addition to SFT-centered pipelines. Code, data, and models are available.
Desiree Cho, Cameron Tice, Bernie Hogan et al.· 0 citations
Lower gestational age and lower birth weight were significantly associated with persistence of PDA, indicating prematurity as a major determinant, highlighting PDA as a prevalent condition contributing significantly to early complications and death.
Avinash Shukla, Amar M Taksande· Journal of Clinical and Diag...· 0 citations
AIMS
Safe prescribing and effective medication review during acute hospitalization depends on accurate information about liver and kidney function because these organs are responsible for the elimination of most medications. While estimates for kidney function are widely used, comparable markers of hepatic drug-metabolizing capacity are not routinely available. We evaluated the clinical relevance of pharmacogenomic (PGx) information for key pharmacogenes implicated in medication elimination in older adults presenting to the emergency department.
METHODS
Fourteen pharmacogenes were analysed using the Personal Medicine Profile™ test. GeneYouIn PillCheck™ software performed genotype-to-phenotype translations, identified drug-gene interactions (DGIs) and generated a clinical decision report based on each patient's actual medication use.
RESULTS
Among 125 acutely admitted older medical patients (median age 78.3 years; 10 medications; 7 chronic diseases), PGx testing identified 88 DGIs across 63 patients (50.4%). Of these, 46.5% were considered by clinical experts to be clinically relevant for the individual patient, affecting 33 patients (26.4%) in the total study population. Frequently implicated pharmacogenes included CYP2C19 (25.0%), SLCO1B1 (25.0%), CYP2D6 (20.5%), CYP2C9 (14.8%) and OPRM1 (6.8%), and frequently implicated medications included losartan (13.6%), pantoprazole (12.5%), simvastatin (12.5%), atorvastatin (11.4%) and metoprolol (11.4%).
CONCLUSION
With more than one-quarter of acutely admitted older medical patients having one or more clinically relevant DGIs, these findings suggest that PGx information may have meaningful clinical utility for improving prescribing practices in acute care. However, interpretation in this population requires careful consideration of other factors such as nutritional status and inflammation that may modify pharmacogene activity.
L. W. Christensen, A. K. Boas, Emilie Clausen et al.· British Journal of Clinical...· 0 citations
The positive results of ISIT counseling on psychological distress symptoms and reducing perinatal grief during follow-ups conducted immediately post-intervention and 3 months after the intervention suggest that it will be useful for managing the complications of spontaneous abortion.
Nazanin Karimihamzekolaee, Hajar Adib-rad, Hajar Pasha et al.· Health Science Reports· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduAug 17, 2026
A USAF cadet and a Lincoln Laboratory researcher found AI chatbots can help nontechnical service members produce viable software applications for their unique problems.
