cs.LGOct 30, 2026

Observable-Specific Validity Tests for Scientific AI: Controlled Approximations and Intervention Response

Authors: Xiaoshn Nee, Haobo Zhong, Xiaomin Ni

Organizations: Independent researcher · HSBC Business School, Peking University, Shenzhen City, Guangdong 518055, China · Artificial Intelligence Research Institute, Shenzhen University of Advanced Technology, Shenzhen, China

Abstract

An approximation can reproduce an energy while missing its response, or satisfy an error tolerance without being certified by a given bound. We formulate observable-specific validity records for scientific artificial intelligence (AI) that link a reduction to its reference model, error measure, tolerance, parameter domain, and transformations or interventions. Explicit displacement and velocity certificates for a singularly perturbed oscillator recover distinct convergence orders and provide sufficient validity guarantees. Exact Hubbard dimer relations resolve the different coupling domains accepted by energy and response criteria. Across 16 repeated neural fits, imposing dimensional homogeneity reduces energy scale transfer error from 52.3% to 0.0768% and raises the median fraction jointly satisfying gap and response tolerances from zero to 100%. Paired transformations verify the protected scaling relation, while a coupling crossover identifies the additional information needed for response transfer. A causal model recovers a positive intervention response beneath a strong negative observational association. For two mechanisms observed with Gaussian noise whose responses first differ at order xkx^k, we derive an exact fixed control query requirement scaling as x−2kx^{-2k} and verify its full success curve numerically. Together with a 256-task evaluation interface, these results provide independently checkable tests of controlled approximation, physical transfer, and mechanism discrimination in scientific AI.

Explore similar work

Jun 11, 2026

Order Is Not Control: Driven-Dissipative Response Laws Across Artificial and Biological Systems

AI alignment, interpretability, steering, and neural perturbation studies identify order-inducing objects. We argue that order is not control. Control requires a receiver-gated response law: a denominator-indexed operator mapping material state, action/drive, bath, and receiver state to response displacement, sinks, effort, and basin projection. We identify it across biological, LLM, adapter, and stochastic-operator panels. The laws are local: an intervention can be admitted, saturated, sign-changing, leaky, or overdriven depending on medium, bath, receiver state, action port, and comparator. Control is assigned when finite effort moves a target or outcome-readout class under the same denominator while damage, null/evasive, invalid format, overdrive, and unnecessary effort stay bounded. Mouse ALM, C. elegans, and zebrafish panels provide physical response-operator evidence while excluding coordinate identity and controller conclusions. LLM panels show generated-output response laws: across four material conditions, response vectors are predictable at 72.8-73.7% component-sign accuracy, rising to 84.3-84.8% on nonzero components; held-out observers predict system-effect and target/oracle families at 93.6% and 91.7% accuracy. Constitution-conditioned adapters reshape susceptibility as prepared media, and stochastic-operator panels separate measured opportunity from deployable action policies. This gives a driven-dissipative response-system account at the mesoscopic control level: drives act through prepared media, baths, and receivers, producing admitted movement, impedance, sinks, or overdrive. The evidence supports local admitted control and measurable stochastic response operators, while leaving deployable pre-generation control, hidden/logit causal sufficiency, biological-to-LLM coordinate identity, and literal thermodynamic quantities outside scope.
Jun 22, 2026

Position: Correct Answer, Wrong Mechanism -- When AI Scientists Defend General Claims Their Own Data Contradicts

AI scientist systems are described as tools, coauthors, or founders, but we evaluate them as if only the final answer matters. This position paper argues that outcome-only evaluation is insufficient, and that task outcome, mechanism fidelity, and epistemic honesty must be measured separately. Our evidence comes from 28 episodes of a coding agent attempting to rediscover a known particle identification observable in a Geant4 simulation, including an 8-episode probe across two additional frontier models. In 4/20 primary-model and 3/8 cross-model episodes, agents reach right-looking results through incorrect reasoning that breaks when conditions change, which we call Correct Answer, Wrong Mechanism (CAWM). Honesty and mechanism fidelity dissociate within a single agent trajectory. When given a partially misleading prior, all five agents reject the false component on evidence, yet one defends its chosen observable with physics inconsistent with its own data. In the simulation-based discovery setting studied here, coding agents prove reliable tools but unreliable scientific co-authors for open-ended claim-making, where co-author trust requires mechanism-fidelity verification they do not reliably self-apply. The failure is detectable, and we propose a lightweight test. A one-step regime-shift check needs only the agent's claim and flags the over-generalized cases. A companion recomputation flags the remaining cases when the correct observable is known. Together, these checks flag every CAWM case in this study.
Mar 26, 2026

A Judge Agent Closes the Reliability Gap in AI-Generated Scientific Simulation

Large language models can generate scientific simulation code, but the generated code silently fails on most non-textbook problems. We show that classical mathematical validation -- well-posedness, convergence, and error certification -- can be fully automated by a Judge Agent, reducing the silent-failure rate from 42% to 1.5% across 134 test cases spanning 12 scientific domains. The headline result comes from a prospective benchmark: 72 blinded tasks submitted by 12 independent scientists yield an 89% success rate (95% CI: [80%, 95%]) with automated error bounds, versus 53% without the Judge. On clinical CT (the only powered experiment, n = 200), the pipeline reaches 99% of expert quality. The residual 1.5% concentrates at bifurcation points where certifiability breaks down. We formalize this boundary through the simulability class S and introduce spec.md, a structured specification format that makes any scientific computation problem machine-readable and solver-independent. Code, data, and all 72 benchmark tasks are publicly archived.