Robert Scholz

Jürgen Gutekunst, Armin Nurkanovic, Ekaterina Kostina et al.

Recently there has been a lot of progress in the development of economic nonlinear model predictive control (NMPC) schemes for multistage optimal power flow (OPF) problems. However, the additional inclusion of discrete decision variables to model generator runtimes and generator startup costs can amount to large scale mixed-integer nonlinear programs (MINLPs) that are computationally very challenging. This work investigates the practical approach that replaces the nonlinear AC power flow equations by convex quadratic approximations. In combination with the discrete generator dynamics this leads to a mixed-integer quadratically constrained program (MIQCP) which is of significantly lower complexity and can be solved in reasonable time by off-the-shelf solvers such as CPLEX. We further show that simple terminal constraints are not sufficient to guarantee recursive feasibility of the NMPC scheme if constraints on generator runtime and on the number of generator startup events are present. To address this challenge we propose the use of additional time-coupled constraints and prove the resulting recursive feasibility property. Based on the assumption of periodic dissipativity of the underlying system we can prove stability of the proposed controller. To illustrate our results, we present simulations of a realistic 6-bus microgrid under different demand scenarios.

Benjamin Warner, Ratna Sagari Grandhi, Max Kieffer et al.

Evaluating large language models (LLMs) for medical applications remains challenging due to benchmark saturation, limited data accessibility, and insufficient coverage of relevant tasks. Existing suites have either saturated, heavily depend on restricted datasets, or lack comprehensive model coverage. We introduce Medmarks, a fully open-source evaluation suite with 30 benchmarks spanning question answering, information extraction, medical calculations, and open-ended clinical reasoning. We perform a systematic evaluation of 61 models across 71 configurations using verifiable metrics and LLM-as-a-Judge. Our results show that frontier reasoning models (Gemini 3 Pro Preview, GPT-5.1, & GPT-5.2) achieve the highest performance across both benchmarks, most frontier proprietary models are significantly more token efficient than open-weight alternatives, medically fine-tuned models outperform their generalist counterparts, and that models are susceptible to answer-order bias (particularly smaller models and Grok 4). A subset of our evals (Medmarks-T) can be directly used as reinforcement learning environments to post-train LLMs for medical reasoning. Code is available at https://github.com/MedARC-AI/Medmarks

Mubashara Akhtar, Anka Reuel, Prajna Soni et al. · meta-ai

Artificial Intelligence (AI) benchmarks play a central role in measuring progress in model development and guiding deployment decisions. However, many benchmarks quickly become saturated, meaning that they can no longer differentiate between the best-performing models, diminishing their long-term value. In this study, we analyze benchmark saturation across 60 Large Language Model (LLM) benchmarks selected from technical reports by major model developers. To identify factors driving saturation, we characterize benchmarks along 14 properties spanning task design, data construction, and evaluation format. We test five hypotheses examining how each property contributes to saturation rates. Our analysis reveals that nearly half of the benchmarks exhibit saturation, with rates increasing as benchmarks age. Notably, hiding test data (i.e., public vs. private) shows no protective effect, while expert-curated benchmarks resist saturation better than crowdsourced ones. Our findings highlight which design choices extend benchmark longevity and inform strategies for more durable evaluation.

Anka Reuel, Avijit Ghosh, Jenny Chim et al.

Foundation models are increasingly central to high-stakes AI systems, and governance frameworks now depend on evaluations to assess their risks and capabilities. Although general capability evaluations are widespread, social impact assessments covering bias, fairness, privacy, environmental costs, and labor practices remain uneven across the AI ecosystem. To characterize this landscape, we conduct the first comprehensive analysis of both first-party and third-party social impact evaluation reporting across a wide range of model developers. Our study examines 186 first-party release reports and 183 post-release evaluation sources, and complements this quantitative analysis with interviews of model developers. We find a clear division of evaluation labor: first-party reporting is sparse, often superficial, and has declined over time in key areas such as environmental impact and bias, while third-party evaluators including academic researchers, nonprofits, and independent organizations provide broader and more rigorous coverage of bias, harmful content, and performance disparities. However, this complementarity has limits. Only model developers can authoritatively report on data provenance, content moderation labor, financial costs, and training infrastructure, yet interviews reveal that these disclosures are often deprioritized unless tied to product adoption or regulatory compliance. Our findings indicate that current evaluation practices leave major gaps in assessing AI's societal impacts, highlighting the urgent need for policies that promote developer transparency, strengthen independent evaluation ecosystems, and create shared infrastructure to aggregate and compare third-party evaluations in a consistent and accessible way.

Robert Scholz, Kunal Bagga, Christine Ahrends et al.

We present our submission to the Algonauts 2025 Challenge, where the goal is to predict fMRI brain responses to movie stimuli. Our approach integrates multimodal representations from large language models, video encoders, audio models, and vision-language models, combining both off-the-shelf and fine-tuned variants. To improve performance, we enhanced textual inputs with detailed transcripts and summaries, and we explored stimulus-tuning and fine-tuning strategies for language and vision models. Predictions from individual models were combined using stacked regression, yielding solid results. Our submission, under the team name Seinfeld, ranked 10th. We make all code and resources publicly available, contributing to ongoing efforts in developing multimodal encoding models for brain activity.