The improvement of economic policymaking presents an opportunity for broad societal benefit, a notion that has inspired research towards AI-driven policymaking tools. AI policymaking holds the potential to surpass human performance through the ability to process data quickly at scale. However, existing RL-based methods exhibit sample inefficiency, and are further limited by an inability to flexibly incorporate nuanced information into their decision-making processes. Thus, we propose a novel method in which we instead utilize pre-trained Large Language Models (LLMs), as sample-efficient policymakers in socially complex multi-agent reinforcement learning (MARL) scenarios. We demonstrate significant efficiency gains, outperforming existing methods across three environments. Our code is available at https://github.com/hegasz/large-legislative-models.

Kidney transplantation is the preferred treatment for end-stage renal disease, yet the scarcity of donors and inefficiencies in allocation systems create major bottlenecks, resulting in prolonged wait times and alarming mortality rates. Despite their severe scarcity, timely and effective interventions to prevent non-utilization of life-saving organs remain inadequate. Expedited out-of-sequence placement of hard-to-place kidneys to centers with the highest likelihood of utilizing them has been recommended in the literature as an effective strategy to improve placement success. Nevertheless, current attempts towards this practice is non-standardized and heavily rely on the subjective judgment of the decision-makers. This paper proposes a novel data-driven, machine learning-based ranking system for allocating hard-to-place kidneys to centers with a higher likelihood of accepting and successfully transplanting them. Using the national deceased donor kidney offer and transplant datasets, we construct a unique dataset with donor-, center-, and patient-specific features. We propose a data-driven out-of-sequence placement policy that utilizes machine learning models to predict the acceptance probability of a given kidney by a set of transplant centers, ranking them accordingly based on their likelihood of acceptance. Our experiments demonstrate that the proposed policy can reduce the average number of centers considered before placement by fourfold for all kidneys and tenfold for hard-to-place kidneys. This significant reduction indicates that our method can improve the utilization of hard-to-place kidneys and accelerate their acceptance, ultimately reducing patient mortality and the risk of graft failure. Further, we utilize machine learning interpretability tools to provide insights into factors influencing the kidney allocation decisions.

The accuracy of density functional theory hinges on the approximation of non-local contributions to the exchange-correlation (XC) functional. To date, machine-learned and human-designed approximations suffer from insufficient accuracy, limited scalability, or dependence on costly reference data. To address these issues, we introduce Equivariant Graph Exchange Correlation (EG-XC), a novel non-local XC functional based on equivariant graph neural networks. EG-XC combines semi-local functionals with a non-local feature density parametrized by an equivariant nuclei-centered point cloud representation of the electron density to capture long-range interactions. By differentiating through a self-consistent field solver, we train EG-XC requiring only energy targets. In our empirical evaluation, we find EG-XC to accurately reconstruct `gold-standard' CCSD(T) energies on MD17. On out-of-distribution conformations of 3BPA, EG-XC reduces the relative MAE by 35% to 50%. Remarkably, EG-XC excels in data efficiency and molecular size extrapolation on QM9, matching force fields trained on 5 times more and larger molecules. On identical training sets, EG-XC yields on average 51% lower MAEs.

There have been many benchmarks for evaluating long-context language models (LCLMs), but developers often rely on synthetic tasks like needle-in-a-haystack (NIAH) or arbitrary subsets of tasks. It remains unclear whether they translate to the diverse downstream applications of LCLMs, and the inconsistency further complicates model comparison. We investigate the underlying reasons behind current practices and find that existing benchmarks often provide noisy signals due to low coverage of applications, insufficient lengths, unreliable metrics, and incompatibility with base models. In this work, we present HELMET (How to Evaluate Long-context Models Effectively and Thoroughly), a comprehensive benchmark encompassing seven diverse, application-centric categories. We also address many issues in previous benchmarks by adding controllable lengths up to 128k tokens, model-based evaluation for reliable metrics, and few-shot prompting for robustly evaluating base models. Consequently, we demonstrate that HELMET offers more reliable and consistent rankings of frontier LCLMs. Through a comprehensive study of 51 LCLMs, we find that (1) synthetic tasks like NIAH are not good predictors of downstream performance; (2) the diverse categories in HELMET exhibit distinct trends and low correlation with each other; and (3) while most LCLMs achieve perfect NIAH scores, open-source models significantly lag behind closed ones when the task requires full-context reasoning or following complex instructions -- the gap widens with increased lengths. Finally, we recommend using our RAG tasks for fast model development, as they are easy to run and more predictive of other downstream performance; ultimately, we advocate for a holistic evaluation across diverse tasks.

Large Language Models (LLMs) exhibit impressive capabilities but require careful alignment with human preferences. Traditional training-time methods finetune LLMs using human preference datasets but incur significant training costs and require repeated training to handle diverse user preferences. Test-time alignment methods address this by using reward models (RMs) to guide frozen LLMs without retraining. However, existing test-time approaches rely on trajectory-level RMs which are designed to evaluate complete responses, making them unsuitable for autoregressive text generation that requires computing next-token rewards from partial responses. To address this, we introduce GenARM, a test-time alignment approach that leverages the Autoregressive Reward Model--a novel reward parametrization designed to predict next-token rewards for efficient and effective autoregressive generation. Theoretically, we demonstrate that this parametrization can provably guide frozen LLMs toward any distribution achievable by traditional RMs within the KL-regularized reinforcement learning framework. Experimental results show that GenARM significantly outperforms prior test-time alignment baselines and matches the performance of training-time methods. Additionally, GenARM enables efficient weak-to-strong guidance, aligning larger LLMs with smaller RMs without the high costs of training larger models. Furthermore, GenARM supports multi-objective alignment, allowing real-time trade-offs between preference dimensions and catering to diverse user preferences without retraining.

Residential rooftop solar adoption is considered crucial for reducing carbon emissions. The lack of photovoltaic (PV) data at a finer resolution (e.g., household, hourly levels) poses a significant roadblock to informed decision-making. We discuss a novel methodology to generate a highly granular, residential-scale realistic dataset for rooftop solar adoption across the contiguous United States. The data-driven methodology consists of: (i) integrated machine learning models to identify PV adopters, (ii) methods to augment the data using explainable AI techniques to glean insights about key features and their interactions, and (iii) methods to generate household-level hourly solar energy output using an analytical model. The resulting synthetic datasets are validated using real-world data and can serve as a digital twin for modeling downstream tasks. Finally, a policy-based case study utilizing the digital twin for Virginia demonstrated increased rooftop solar adoption with the 30\% Federal Solar Investment Tax Credit, especially in Low-to-Moderate-Income communities.

Gradient-based interpretations often require an anchor point of comparison to avoid saturation in computing feature importance. We show that current baselines defined using static functions--constant mapping, averaging or blurring--inject harmful colour, texture or frequency assumptions that deviate from model behaviour. This leads to accumulation of irregular gradients, resulting in attribution maps that are biased, fragile and manipulable. Departing from the static approach, we propose UNI to compute an (un)learnable, debiased and adaptive baseline by perturbing the input towards an unlearning direction of steepest ascent. Our method discovers reliable baselines and succeeds in erasing salient features, which in turn locally smooths the high-curvature decision boundaries. Our analyses point to unlearning as a promising avenue for generating faithful, efficient and robust interpretations.

Recent advancements in test time compute, particularly through the use of verifier models, have significantly enhanced the reasoning capabilities of Large Language Models (LLMs). This generator-verifier approach closely resembles the actor-critic framework in reinforcement learning (RL). However, current verifier models in LLMs often rely on supervised fine-tuning without temporal difference learning such as Q-learning. This paper introduces VerifierQ, a novel approach that integrates Offline Q-learning into LLM verifier models. We address three key challenges in applying Q-learning to LLMs: (1) handling utterance-level Markov Decision Processes (MDPs), (2) managing large action spaces, and (3) mitigating overestimation bias. VerifierQ introduces a modified Bellman update for bounded Q-values, incorporates Implicit Q-learning (IQL) for efficient action space management, and integrates a novel Conservative Q-learning (CQL) formulation for balanced Q-value estimation. Our method enables parallel Q-value computation and improving training efficiency. While recent work has explored RL techniques like MCTS for generators, VerifierQ is among the first to investigate the verifier (critic) aspect in LLMs through Q-learning. This integration of RL principles into verifier models complements existing advancements in generator techniques, potentially enabling more robust and adaptive reasoning in LLMs. Experimental results on mathematical reasoning tasks demonstrate VerifierQ's superior performance compared to traditional supervised fine-tuning approaches, with improvements in efficiency, accuracy and robustness. By enhancing the synergy between generation and evaluation capabilities, VerifierQ contributes to the ongoing evolution of AI systems in addressing complex cognitive tasks across various domains.

The rise of AI-generated content in popular information sources raises significant concerns about accountability, accuracy, and bias amplification. Beyond directly impacting consumers, the widespread presence of this content poses questions for the long-term viability of training language models on vast internet sweeps. We use GPTZero, a proprietary AI detector, and Binoculars, an open-source alternative, to establish lower bounds on the presence of AI-generated content in recently created Wikipedia pages. Both detectors reveal a marked increase in AI-generated content in recent pages compared to those from before the release of GPT-3.5. With thresholds calibrated to achieve a 1% false positive rate on pre-GPT-3.5 articles, detectors flag over 5% of newly created English Wikipedia articles as AI-generated, with lower percentages for German, French, and Italian articles. Flagged Wikipedia articles are typically of lower quality and are often self-promotional or partial towards a specific viewpoint on controversial topics.

We propose a new variant of the strategic classification problem: a principal reveals a classifier, and $n$ agents report their (possibly manipulated) features to be classified. Motivated by real-world applications, our model crucially allows the manipulation of one agent to affect another; that is, it explicitly captures inter-agent externalities. The principal-agent interactions are formally modeled as a Stackelberg game, with the resulting agent manipulation dynamics captured as a simultaneous game. We show that under certain assumptions, the pure Nash Equilibrium of this agent manipulation game is unique and can be efficiently computed. Leveraging this result, PAC learning guarantees are established for the learner: informally, we show that it is possible to learn classifiers that minimize loss on the distribution, even when a random number of agents are manipulating their way to a pure Nash Equilibrium. We also comment on the optimization of such classifiers through gradient-based approaches. This work sets the theoretical foundations for a more realistic analysis of classifiers that are robust against multiple strategic actors interacting in a common environment.