Scientific reasoning, the process through which humans apply logic, evidence, and critical thinking to explore and interpret scientific phenomena, is essential in advancing knowledge reasoning across diverse fields. However, despite significant progress, current scientific reasoning models still struggle with generalization across domains and often fall short of multimodal perception. Multimodal Large Language Models (MLLMs), which integrate text, images, and other modalities, present an exciting opportunity to overcome these limitations and enhance scientific reasoning. Therefore, this position paper argues that MLLMs can significantly advance scientific reasoning across disciplines such as mathematics, physics, chemistry, and biology. First, we propose a four-stage research roadmap of scientific reasoning capabilities, and highlight the current state of MLLM applications in scientific reasoning, noting their ability to integrate and reason over diverse data types. Second, we summarize the key challenges that remain obstacles to achieving MLLM's full potential. To address these challenges, we propose actionable insights and suggestions for the future. Overall, our work offers a novel perspective on MLLM integration with scientific reasoning, providing the LLM community with a valuable vision for achieving Artificial General Intelligence (AGI).

Abstract Mass spectrometry (MS) has revealed millions of small organic molecules across organisms, yet most remain uncharacterized, limiting progress in biology and medicine. Despite computational advances, MS workflows rely heavily on expert input and reference libraries that cover only a fraction of known chemical space. Here, we introduce AIMe (AI Molecule Explorer), a multi-agent neuro-symbolic AI framework that transforms the interpretation of unknown spectra into an omics-scale exploration across the known structural space, providing chemically interpretable annotations. At its core, AIMe combines chemical reasoning with structure- informed learning to predict MS 2 spectra by modeling fragmentation as a sequence of actions, outperforming existing methods. AIMe dynamically constructs fragmentation pathways by assigning likelihoods to individual fragmentation actions, linking spectral peaks to explicit fragment molecular formulas and structures. At scale, AIMe predicted MS 2 spectra for over 100 million small organic molecules in PubChem and organized them into MS 2 KOSMOS, a substructure-informed community resource comprising over 800 million predicted spectra that expands the searchable small-molecule universe by roughly three orders of magnitude relative to experimental libraries. Analogous to sequence homology-based searches in genomics and proteomics, AIMe maps unknown spectra to molecular neighborhoods in MS 2 KOSMOS. Exact- formula indexing enables ranked retrieval of candidates and related structures, with peak-level structural and fragmentation-pathway annotations. Applied to mouse microbiota-dependent metabolites, AIMe enabled putative annotation of knowns and guided structure elucidation of unknowns, revealing previously unreported types of microbiota-dependent polyamines that also occur in humans. At repository scale, AIMe enabled putative annotation of roughly a third of 7 million spectral clusters representing most of the unknowns in the GNPS database. By extending MS 2 annotation beyond curated-library matching to interpretable search across the known small-molecule universe, AIMe accelerates discovery and large-scale exploration of small molecules across biomedicine, agriculture, and ecology.

We study mathematical discovery through the lens of neurosymbolic reasoning, where an AI agent powered by a large language model (LLM), coupled with symbolic computation tools, and human strategic direction, jointly produced a new result in combinatorial design theory. The main result of this human-AI collaboration is a tight lower bound on the imbalance of Latin squares for the notoriously difficult case n ≡ 1 3. We reconstruct the discovery process from detailed interaction logs spanning multiple sessions over several days and identify the distinct cognitive contributions of each component. The AI agent proved effective at uncovering hidden structure and generating hypotheses. The symbolic component consists of computer algebra, constraint solvers, and simulated annealing, which provides rigorous verification and exhaustive enumeration. Human steering supplied the critical research pivot that transformed a dead end into a productive inquiry. Our analysis reveals that multi-model deliberation among frontier LLMs proved reliable for criticism and error detection but unreliable for constructive claims. The resulting human-AI mathematical contribution, a tight lower bound of 4n(n-1)/9, is achieved via a novel class of near-perfect permutations. The bound was formally verified in Lean 4. Our experiments show that neurosymbolic systems can indeed produce genuine discoveries in pure mathematics.

Neural solvers have shown significant potential in solving the Traveling Salesman Problem (TSP), yet current approaches face significant challenges. Supervised learning (SL)-based solvers require large amounts of high-quality labeled data, while reinforcement learning (RL)-based solvers, though less dependent on such data, often suffer from inefficiencies. To address these limitations, we propose LocalEscaper, a novel weakly-supervised learning framework for large-scale TSP. LocalEscaper effectively combines the advantages of both SL and RL, enabling effective training on datasets with low-quality labels. To further enhance solution quality, we introduce a regional reconstruction strategy, which is the key technique of this paper and mitigates the local-optima problem common in existing local reconstruction methods. Experimental results on both synthetic and real-world datasets demonstrate that LocalEscaper outperforms existing neural solvers, achieving remarkable results.

Nature-inspired computation is receiving increasing attention. Various Ising machine (IM) implementations have recently been proven to be effective in solving numerous combinatorial optimization problems including maximum cut, low density parity check (LDPC) decoding, and Boolean satisfiability (SAT) problems. In this paper, a novel method is presented to solve SAT or MAX-SAT problems with a CMOS circuit implementation. The technique solves a SAT problem by mapping the SAT variables onto quantized capacitor voltages generated by an array of nodes that interact through a network of coupling units. The nodal interaction is achieved through coupling currents produced by the coupling units, which charge or discharge capacitor voltages, implementing a gradient descent along the SAT problem's cost function to minimize the number of unsatisfied clauses. The system also incorporates a unique low-complexity perturbation scheme to avoid settling in local minima, greatly enhancing the performance of the system. The simulation results demonstrate that the proposed SKI-SAT is a high-performance and low-energy alternative that surpasses existing software-based SAT solvers by significant margins, achieving more than 10 times faster solution and over 300 times less power.

Much of the complexity of social, biological, and engineered systems arises from a network of complex interactions connecting many basic components. Network analysis tools have been successful at uncovering latent structure termed communities in such networks. However, some of the most interesting structure can be difficult to uncover because it is obscured by the more dominant structure. Our previous work proposes a general structure amplification technique called HICODE that uncovers many layers of functional hidden structure in complex networks. HICODE incrementally weakens dominant structure through randomization allowing the hidden functionality to emerge, and uncovers these hidden structure in real-world networks that previous methods rarely uncover. In this work, we conduct a comprehensive and systematic theoretical analysis on the hidden community structure. In what follows, we define multi-layer stochastic block model, and provide theoretical support using the model on why the existence of hidden structure will make the detection of dominant structure harder compared with equivalent random noise. We then provide theoretical proofs that the iterative reducing methods could help promote the uncovering of hidden structure as well as boosting the detection quality of dominant structure.

Large-scale AI systems that combine search and learning have reached super-human levels of performance in game-playing, but have also been shown to fail in surprising ways. The brittleness of such models limits their efficacy and trustworthiness in real-world deployments. In this work, we systematically study one such algorithm, AlphaZero, and identify two phenomena related to the nature of exploration. First, we find evidence of policy-value misalignment -- for many states, AlphaZero's policy and value predictions contradict each other, revealing a tension between accurate move-selection and value estimation in AlphaZero's objective. Further, we find inconsistency within AlphaZero's value function, which causes it to generalize poorly, despite its policy playing an optimal strategy. From these insights we derive VISA-VIS: a novel method that improves policy-value alignment and value robustness in AlphaZero. Experimentally, we show that our method reduces policy-value misalignment by up to 76%, reduces value generalization error by up to 50%, and reduces average value error by up to 55%.

The top-$k$ classification accuracy is a crucial metric in machine learning and is often used to evaluate the performance of deep neural networks. These networks are typically trained using the cross-entropy loss, which optimizes for top-$1$ classification and is considered optimal in the case of infinite data. However, in real-world scenarios, data is often noisy and limited, leading to the need for more robust losses. In this paper, we propose using the Weighted Sampling Without Replacement (WSWR) method as a learning objective for top-$k$ loss. While traditional methods for evaluating **WSWR-based top-$k$ loss** are computationally impractical, we show a novel connection between WSWR and Reinforcement Learning (RL) and apply well-established RL algorithms to estimate gradients. We compared our method with recently proposed top-$k$ losses in various regimes of noise and data size for the prevalent use case of $k = 5$. Our experimental results reveal that our method consistently outperforms all other methods on the top-$k$ metric for noisy datasets, has more robustness on extreme testing scenarios, and achieves competitive results on training with limited data.

Despite the success of practical solvers in various NP-complete domains such as SAT and CSP as well as using deep reinforcement learning to tackle two-player games such as Go, certain classes of PSPACE-hard planning problems have remained out of reach. Even carefully designed domain-specialized solvers can fail quickly due to the exponential search space on hard instances. Recent works that combine traditional search methods, such as best-first search and Monte Carlo tree search, with Deep Neural Networks' (DNN) heuristics have shown promising progress and can solve a significant number of hard planning instances beyond specialized solvers. To better understand why these approaches work, we studied the interplay of the policy and value networks of DNN-based best-first search on Sokoban and show the surprising effectiveness of the policy network, further enhanced by the value network, as a guiding heuristic for the search. To further understand the phenomena, we studied the cost distribution of the search algorithms and found that Sokoban instances can have heavy-tailed runtime distributions, with tails both on the left and right-hand sides. In particular, for the first time, we show the existence of \textit{left heavy tails} and propose an abstract tree model that can empirically explain the appearance of these tails. The experiments show the critical role of the policy network as a powerful heuristic guiding the search, which can lead to left heavy tails with polynomial scaling by avoiding exploring exponentially sized subtrees. Our results also demonstrate the importance of random restarts, as are widely used in traditional combinatorial solvers, for DNN-based search methods to avoid left and right heavy tails.

We study fairness through the lens of cooperative multi-agent learning. Our work is motivated by empirical evidence that naive maximization of team reward yields unfair outcomes for individual team members. To address fairness in multi-agent contexts, we introduce team fairness, a group-based fairness measure for multi-agent learning. We then prove that it is possible to enforce team fairness during policy optimization by transforming the team's joint policy into an equivariant map. We refer to our multi-agent learning strategy as Fairness through Equivariance (Fair-E) and demonstrate its effectiveness empirically. We then introduce Fairness through Equivariance Regularization (Fair-ER) as a soft-constraint version of Fair-E and show that it reaches higher levels of utility than Fair-E and fairer outcomes than non-equivariant policies. Finally, we present novel findings regarding the fairness-utility trade-off in multi-agent settings; showing that the magnitude of the trade-off is dependent on agent skill.

Despite the tremendous success of applying traditional backtrack-style combinatorial search methods in various NP-complete domains such as SAT and CSP as well as using deep reinforcement learning (RL) to tackle two-player games such as Go, PSPACE-hard AI planning has remained out of reach for current AI planning systems. Even carefully designed domain-specific solvers fail quickly due to the exponential combinatorial search space on hard instances. Recent work based on deep learning guided search algorithms that combine traditional search-based methods, such as A\textsc{*} and MCTS search, with deep neural networks' heuristic prediction has shown promising progress. These methods can solve a significant number of hard planning instances beyond specialized solvers. To better understanding why these approaches work we study the interplay of the policy and value networks in A\textsc{*}-based deep RL and show the surprising effectiveness of the policy network, further enhanced by the value network, as a guiding heuristic for A\textsc{*}. To further understand the phenomena, we study the cost distributions of deep planners and found planning instances can have heavy-tailed runtime distributions, with tails both on the right-hand and left-hand sides. In particular, for the first time, we show the existence of {\textit{left}} heavy tails and propose a theoretical model that can explain the appearance of these tails. We provide extensive experimental data supporting our model. The experiments show the critical role of the policy network as a powerful heuristic guiding A\textsc{*}, which can lead to left tails with polynomial scaling by avoiding exploring exponential size sub-trees early on in the search. Our results also demonstrate the importance of random restart strategies, as are widely used in traditional combinatorial solvers, for deep reinforcement learning and deep AI planning systems to avoid left and right heavy tails.

Flesh encodes a variety of haptic information including deformation, temperature, vibration, and damage stimuli using a multisensory array of mechanoreceptors distributed on the surface of the human body. Currently, soft sensors are capable of detecting some haptic stimuli, but whole-body multimodal perception at scales similar to a human adult (surface area ~17,000 square centimeters) is still a challenge in artificially intelligent agents due to the lack of encoding. This encoding is needed to reduce the wiring required to send the vast amount of information transmitted to the processor. We created a robotic flesh that could be further developed for use in these agents. This engineered flesh is an optical, elastomeric matrix “innervated” with stretchable lightguides that encodes haptic stimuli into light: temperature into wavelength due to thermochromic dyes and forces into intensity due to mechanical deformation. By exploiting the optical properties of the constitutive materials and using machine learning, we infer spatiotemporal, haptic information from light that is read by an image sensor. We demonstrate the capabilities of our system in various assemblies to estimate temperature, contact location, normal and shear force, gestures, and damage from temporal snapshots of light coming from the entire haptic sensor with errors <5%.

In recent years, deep Reinforcement Learning (RL) has been successful in various combinatorial search domains, such as two-player games and scientific discovery. However, directly applying deep RL in planning domains is still challenging. One major difficulty is that without a human-crafted heuristic function, reward signals remain zero unless the learning framework discovers any solution plan. Search space becomes \emph{exponentially larger} as the minimum length of plans grows, which is a serious limitation for planning instances with a minimum plan length of hundreds to thousands of steps. Previous learning frameworks that augment graph search with deep neural networks and extra generated subgoals have achieved success in various challenging planning domains. However, generating useful subgoals requires extensive domain knowledge. We propose a domain-independent method that augments graph search with graph value iteration to solve hard planning instances that are out of reach for domain-specialized solvers. In particular, instead of receiving learning signals only from discovered plans, our approach also learns from failed search attempts where no goal state has been reached. The graph value iteration component can exploit the graph structure of local search space and provide more informative learning signals. We also show how we use a curriculum strategy to smooth the learning process and perform a full analysis of how graph value iteration scales and enables learning.

Emergent communication has made strides towards learning communication from scratch, but has focused primarily on protocols that resemble human language. In nature, multi-agent cooperation gives rise to a wide range of communication that varies in structure and complexity. In this work, we recognize the full spectrum of communication that exists in nature and propose studying lower-level communication. Specifically, we study emergent implicit signaling in the context of decentralized multi-agent learning in difficult, sparse reward environments. However, learning to coordinate in such environments is challenging. We propose a curriculum-driven strategy that combines: (i) velocity-based environment shaping, tailored to the skill level of the multi-agent team; and (ii) a behavioral curriculum that helps agents learn successful single-agent behaviors as a precursor to learning multi-agent behaviors. Pursuit-evasion experiments show that our approach learns effective coordination, significantly outperforming sophisticated analytical and learned policies. Our method completes the pursuit-evasion task even when pursuers move at half of the evader's speed, whereas the highest-performing baseline fails at 80% of the evader's speed. Moreover, we examine the use of implicit signals in coordination through position-based social influence. We show that pursuers trained with our strategy exchange more than twice as much information (in bits) than baseline methods, indicating that our method has learned, and relies heavily on, the exchange of implicit signals.

A new wave of innovation disrupts our social and economic life as we have lived it in the 20th century. On the one hand, artificial intelligence (AI) has the potential to increase our efficiency and ability to cope with the changes that are coming. On the other hand, AI itself also causes disruptions that may be managed best by contributions from other fields. We aim to start an interdisciplinary discourse with this special issue and bring together a set of articles from AI researchers, sociologists, economists, and psychologists who discuss five basic questions: How can AI help? How will AI affect economy and work? How does AI affect democracy? How does AI affect our togetherness? And how does AI affect the human self?

We study fairness through the lens of cooperative multi-agent learning. Our work is motivated by empirical evidence that naive maximization of team reward yields unfair outcomes for individual team members. To address fairness in multi-agent contexts, we introduce team fairness, a group-based fairness measure for multi-agent learning. We then incorporate team fairness into policy optimization -- introducing Fairness through Equivariance (Fair-E), a novel learning strategy that achieves provably fair reward distributions. We then introduce Fairness through Equivariance Regularization (Fair-ER) as a soft-constraint version of Fair-E and show that Fair-ER reaches higher levels of utility than Fair-E and fairer outcomes than policies with no equivariance. Finally, we investigate the fairness-utility trade-off in multi-agent settings.

We propose a curriculum-driven learning strategy for solving difficult multi-agent coordination tasks. Our method is inspired by a study of animal communication, which shows that two straightforward design features (mutual reward and decentralization) support a vast spectrum of communication protocols in nature. We highlight the importance of similarly interpreting emergent communication as a spectrum. We introduce a toroidal, continuous-space pursuit-evasion environment and show that naive decentralized learning does not perform well. We then propose a novel curriculum-driven strategy for multi-agent learning. Experiments with pursuit-evasion show that our approach enables decentralized pursuers to learn to coordinate and capture a superior evader, significantly outperforming sophisticated analytical policies. We argue through additional quantitative analysis -- including influence-based measures such as Instantaneous Coordination -- that emergent implicit communication plays a large role in enabling superior levels of coordination.

Crystal-structure phase mapping is a core, long-standing challenge in materials science that requires identifying crystal structures, or mixtures thereof, in synthesized materials. Materials science experts excel at solving simple systems but cannot solve complex systems, creating a major bottleneck in high-throughput materials discovery. Herein we show how to automate crystal-structure phase mapping. We formulate phase mapping as an unsupervised pattern demixing problem and describe how to solve it using Deep Reasoning Networks (DRNets). DRNets combine deep learning with constraint reasoning for incorporating scientific prior knowledge and consequently require only a modest amount of (unlabeled) data. DRNets compensate for the limited data by exploiting and magnifying the rich prior knowledge about the thermodynamic rules governing the mixtures of crystals with constraint reasoning seamlessly integrated into neural network optimization. DRNets are designed with an interpretable latent space for encoding prior-knowledge domain constraints and seamlessly integrate constraint reasoning into neural network optimization. DRNets surpass previous approaches on crystal-structure phase mapping, unraveling the Bi-Cu-V oxide phase diagram, and aiding the discovery of solar-fuels materials.

Model counting, or counting the number of solutions of a propositional formula, generalizes SAT and is the canonical #P-complete problem. Surprisingly, model counting is hard even for some polynomial-time solvable cases like 2-SAT and Horn-SAT. Efficient algorithms for this problem will have a significant impact on many application areas that are inherently beyond SAT, such as bounded-length adversarial and contingency planning, and, perhaps most importantly, general probabilistic inference. Model counting can be solved, in principle and to an extent in practice, by extending the two most successful frameworks for SAT algorithms, namely, DPLL and local search. However, scalability and accuracy pose a substantial challenge. As a result, several new ideas have been introduced in the last few years that go beyond the techniques usually employed in most SAT solvers. These include division into components, caching, compilation into normal forms, exploitation of solution sampling methods, and certain randomized streamlining techniques using special constraints. This chapter discusses these techniques, exploring both exact methods as well as fast estimation approaches, including those that provide probabilistic or statistical guarantees on the quality of the reported lower or upper bound on the model count.

A new wave of innovation disrupts our social and economic life as we have lived it in the 20th century. On the one hand, artificial intelligence (AI) has the potential to increase our efficiency and ability to cope with the changes that are coming. On the other hand, AI itself also causes disruptions that may be managed best by contributions from other fields. We aim to start an interdisciplinary discourse with this special issue and bring together a set of articles from AI researchers, sociologists, economists, and psychologists who discuss five basic questions: How can AI help? How will AI affect economy and work? How does AI affect democracy? How does AI affect our togetherness? And how does AI affect the human self?