PHREEQC-MCQ-200: A Diagnostic Benchmark for Tool-Augmented Scientific Simulator Agents

Large language model agents are increasingly connected to scientific software, yet it remains unclear when tool access makes scientific computation more reliable rather than merely more complex. We introduce PHREEQC-MCQ-200, a benchmark for evaluating tool-augmented agents on deterministic aqueous-geochemistry simulations. The benchmark contains 200 multiple-choice questions derived from 21 validated PHREEQC scenarios, requiring agents to construct simulator inputs, execute PHREEQC, inspect structured outputs, and commit to final answers. Across multiple frontier and mid-tier model families, simulator access substantially improves aggregate accuracy, confirming that grounded execution is necessary for many scientific-computation tasks. However, the gains are not monotonic: tool-augmented agents also lose items they answered correctly without tools, revealing regressions that average accuracy alone hides. We further show that output-access protocol matters. A table-of-contents interface can reduce token cost while preserving or improving accuracy for stronger models, but it degrades performance for mid-tier models that cannot reliably navigate structured simulator outputs. PHREEQC-MCQ-200 therefore frames scientific tool use as an end-to-end diagnostic problem rather than a simple tool-calling capability. We argue that evaluations of scientific agents should report not only accuracy, but also item-level retention, output-access sensitivity, trajectory failures, and where the computation chain breaks.

Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization

Theorem-proving benchmarks evaluate proof search against fixed formal statements, but natural-language-to-Lean formalization must generate the formal statement itself. In this setting, compilation is only a validity check: a Lean declaration may type-check while omitting hypotheses, changing domains, or expressing a vacuous claim. We study faithful statement formalization as both an evaluation problem and a bottleneck-attribution problem. On a 400-entry graduate-level benchmark spanning real analysis, complex analysis, topology, and algebra, our protocol combines Lean compilation, cross-model semantic judging, and human expert calibration. The resulting picture is different from compile-rate evaluation: a full tool-augmented agent reaches 89.5% compilation but only 60.5% consensus faithfulness, exposing a 29.0-point compile-pass but consensus-unfaithful gap. Targeted human audits support the metric as a conservative decision boundary: across available case-level audits, 96.0% of consensus-positive outputs are human-confirmed faithful, while 82.4% of compile-pass consensus-negative outputs are human-confirmed semantic failures. Under this metric, existing one-shot formalizer models and prover-oriented Lean models remain low, suggesting that formal validity, proof-oriented Lean competence, and faithful statement generation should be reported separately. We then use a full $2^3$ factorial design to decompose three recurring interventions in formalization pipelines: parametric expert drafting, Mathlib/context search, and Lean elaboration feedback. Elaboration feedback is the largest validity intervention, but it also exposes a larger compile-pass semantic-failure bucket; search mainly improves grounding and selectivity; and fine-tuned drafting is largely substitutable in this tool stack once feedback and grounding are available.

Test-Driven Agentic Framework for Reliable Robot Controller

In this work, we present a test-driven, agentic framework for synthesizing a deployable low-level robot controller for navigation tasks. Given a 2D map with an image of an ultrasonic sensor-based robot, or a 3D robotic simulation environment, our framework iteratively refines the generated controller code using diagnostic feedback from structured test suites to achieve task success. We propose a dual-tier repair strategy to refine the generated code that alternates between prompt-level refinement and direct code editing. We evaluate the approach across 2D navigation tasks and 3D navigation in the Webots simulator. Experimental results show that test-driven synthesis substantially improves controller reliability and robustness over one-shot controller generation, especially when the initial prompt is underspecified. The source code and demonstration videos are available at: https://shivanshutripath.github.io/robotic_controller.github.io.

PUNCH: Physics-informed Uncertainty-aware Network for Coronary Hemodynamics

Coronary microvascular dysfunction (CMD) affects millions worldwide yet remains underdiagnosed because gold-standard physiological measurements are invasive and variably reproducible. We introduce a non-invasive, uncertainty-aware framework for estimating coronary flow reserve (CFR) directly from standard angiography. The system integrates physics-informed neural networks with variational inference to infer coronary blood flow from first-principles models of contrast transport, without requiring ground-truth flow measurements. The pipeline runs in approximately three minutes per patient on a single GPU, with no population-level training. Using 1{,}000 synthetic spatiotemporal intensity maps (kymographs) with controlled noise and artifacts, the framework reliably identifies degraded data and outputs appropriately inflated uncertainty estimates, showing strong correspondence between predictive uncertainty and error (Pearson $r = 0.997$, Spearman $ρ= 0.998$). Clinical validation in 12 patients shows strong agreement between PUNCH-derived CFR and invasive bolus thermodilution (Pearson $r = 0.90$, $p = 6.3 \times 10^{-5}$). We focus on the LAD, the artery most commonly assessed in routine CMD testing. Probabilistic CFR estimates have confidence intervals narrower than the variability of repeated invasive measurements. By transforming routine angiography into quantitative, uncertainty-aware assessment, this approach enables scalable, safer, and more reproducible evaluation of coronary microvascular function. Because standard angiography is widely available globally, the framework could expand access to CMD diagnosis and establish a new paradigm for physics-informed, patient-specific inference from clinical imaging.

NewPINNs: Physics-Informing Neural Networks Using Conventional Solvers for Partial Differential Equations

We introduce NewPINNs, a physics-informing learning framework that couples neural networks with conventional numerical solvers for solving differential equations. Rather than enforcing governing equations and boundary conditions through residual-based loss terms, NewPINNs integrates the solver directly into the training loop and defines learning objectives through solver-consistency. The neural network produces candidate solution states that are advanced by the numerical solver, and training minimizes the discrepancy between the network prediction and the solver-evolved state. This pull-push interaction enables the network to learn physically admissible solutions through repeated exposure to the solver's action, without requiring problem-specific loss engineering or explicit evaluation of differential equation residuals. By delegating the enforcement of physics, boundary conditions, and numerical stability to established numerical solvers, NewPINNs mitigates several well-known failure modes of standard physics-informed neural networks, including optimization pathologies, sensitivity to loss weighting, and poor performance in stiff or nonlinear regimes. We demonstrate the effectiveness of the proposed approach across multiple forward and inverse problems involving finite volume, finite element, and spectral solvers.

Data-Efficient Physics-Informed Learning to Model Synchro-Waveform Dynamics of Grid-Integrated Inverter-Based Resources

Inverter-based resources (IBRs) exhibit fast transient dynamics during network disturbances, which often cannot be properly captured by phasor and SCADA measurements. This shortcoming has recently been addressed with the advent of waveform measurement units (WMUs), which provide high-resolution, time-synchronized raw voltage and current waveform samples from multiple locations in the power system. However, transient model learning based on synchro-waveform measurements remains constrained by the scarcity of network disturbances and the complexity of the underlying nonlinear dynamics of IBRs. We propose to address these problems by developing a data-efficient physics-informed machine learning (PIML) framework for synchro-waveform analytics that estimates the IBR terminal current response from only a few network disturbance signatures. Here, the physics of the electrical circuits are used to compensate for limited data availability by constraining the learning process through known circuit relationships. Two cases are considered, with known and unknown circuit parameters. In the latter case, the framework jointly learns the transient dynamics of the IBRs and the parameters of the electrical circuit. Case studies using WMU disturbance data across multiple sampling rates shows consistently lower current estimation error with substantially fewer training events than a purely data-driven baseline.

Aligning to What? Limits to RLHF Based Alignment

Reinforcement Learning from Human Feedback (RLHF) is increasingly used to align large language models (LLMs) with human preferences. However, the effectiveness of RLHF in addressing underlying biases remains unclear. This study investigates the relationship between RLHF and both covert and overt biases in LLMs, particularly focusing on biases against African Americans. We applied various RLHF techniques (DPO, ORPO, and RLOO) to Llama 3 8B and evaluated the covert and overt biases of the resulting models using matched-guise probing and explicit bias testing. We performed additional tests with DPO on different base models and datasets; among several implications, we found that SFT before RLHF calcifies model biases. Additionally, we extend the tools for measuring biases to multi-modal models. Through our experiments we collect evidence that indicates that current alignment techniques are inadequate for nebulous tasks such as mitigating covert biases, highlighting the need for capable datasets, data curating techniques, or alignment tools.

Fine Tuning without Catastrophic Forgetting via Selective Low Rank Adaptation

Adapting deep learning models to new domains often requires computationally intensive retraining and risks catastrophic forgetting. While fine-tuning enables domain-specific adaptation, it can reduce robustness to distribution shifts, impacting out-of-distribution (OOD) performance. Pre-trained zero-shot models like CLIP offer strong generalization but may suffer degraded robustness after fine-tuning. Building on Task Adaptive Parameter Sharing (TAPS), we propose a simple yet effective extension as a parameter-efficient fine-tuning (PEFT) method, using an indicator function to selectively activate Low-Rank Adaptation (LoRA) blocks. Our approach minimizes knowledge loss, retains its generalization strengths under domain shifts, and significantly reduces computational costs compared to traditional fine-tuning. We demonstrate that effective fine-tuning can be achieved with as few as 5\% of active blocks, substantially improving efficiency. Evaluations on pre-trained models such as CLIP and DINO-ViT demonstrate our method's broad applicability and effectiveness in maintaining performance and knowledge retention.

Where Did Your Model Learn That? Label-free Influence for Self-supervised Learning

Self-supervised learning (SSL) has revolutionized learning from large-scale unlabeled datasets, yet the intrinsic relationship between pretraining data and the learned representations remains poorly understood. Traditional supervised learning benefits from gradient-based data attribution tools like influence functions that measure the contribution of an individual data point to model predictions. However, existing definitions of influence rely on labels, making them unsuitable for SSL settings. We address this gap by introducing Influence-SSL, a novel and label-free approach for defining influence functions tailored to SSL. Our method harnesses the stability of learned representations against data augmentations to identify training examples that help explain model predictions. We provide both theoretical foundations and empirical evidence to show the utility of Influence-SSL in analyzing pre-trained SSL models. Our analysis reveals notable differences in how SSL models respond to influential data compared to supervised models. Finally, we validate the effectiveness of Influence-SSL through applications in duplicate detection, outlier identification and fairness analysis. Code is available at: \url{https://github.com/cryptonymous9/Influence-SSL}.

Physics-Informed Neural Networks and Extensions

In this paper, we review the new method Physics-Informed Neural Networks (PINNs) that has become the main pillar in scientific machine learning, we present recent practical extensions, and provide a specific example in data-driven discovery of governing differential equations.