Paul Sajda

Zhengran Ji, Lingyu Zhang, Paul Sajda et al.

Learning collaborative behaviors is essential for multi-agent systems. Traditionally, multi-agent reinforcement learning solves this implicitly through a joint reward and centralized observations, assuming collaborative behavior will emerge. Other studies propose to learn from demonstrations of a group of collaborative experts. Instead, we propose an efficient and explicit way of learning collaborative behaviors in multi-agent systems by leveraging expertise from only a single human. Our insight is that humans can naturally take on various roles in a team. We show that agents can effectively learn to collaborate by allowing a human operator to dynamically switch between controlling agents for a short period and incorporating a human-like theory-of-mind model of teammates. Our experiments showed that our method improves the success rate of a challenging collaborative hide-and-seek task by up to 58% with only 40 minutes of human guidance. We further demonstrate our findings transfer to the real world by conducting multi-robot experiments.

Sharath Koorathota, Nikolas Papadopoulos, Jia Li Ma et al.

Vision Transformers (ViT) have advanced computer vision, yet their efficacy in complex tasks like driving remains less explored. This study enhances ViT by integrating human eye gaze, captured via eye-tracking, to increase prediction accuracy in driving scenarios under uncertainty in both real-world and virtual reality scenarios. First, we establish the significance of human eye gaze in left-right driving decisions, as observed in both human subjects and a ViT model. By comparing the similarity between human fixation maps and ViT attention weights, we reveal the dynamics of overlap across individual heads and layers. This overlap demonstrates that fixation data can guide the model in distributing its attention weights more effectively. We introduce the fixation-attention intersection (FAX) loss, a novel loss function that significantly improves ViT performance under high uncertainty conditions. Our results show that ViT, when trained with FAX loss, aligns its attention with human gaze patterns. This gaze-informed approach has significant potential for driver behavior analysis, as well as broader applications in human-centered AI systems, extending ViT's use to complex visual environments.

Xiaoxiao Sun, Paul Sajda

There is a growing interest in characterizing circular data found in biological systems. Such data are wide ranging and varied, from signal phase in neural recordings to nucleotide sequences in round genomes. Traditional clustering algorithms are often inadequate due to their limited ability to distinguish differences in the periodic component. Current clustering schemes that work in a polar coordinate system have limitations, such as being only angle-focused or lacking generality. To overcome these limitations, we propose a new analysis framework that utilizes projections onto a cylindrical coordinate system to better represent objects in a polar coordinate system. Using the mathematical properties of circular data, we show our approach always finds the correct clustering result within the reconstructed dataset, given sufficient periodic repetitions of the data. Our approach is generally applicable and adaptable and can be incorporated into most state-of-the-art clustering algorithms. We demonstrate on synthetic and real data that our method generates more appropriate and consistent clustering results compared to standard methods. In summary, our proposed analysis framework overcomes the limitations of existing polar coordinate-based clustering methods and provides a more accurate and efficient way to cluster circular data.

Xueqing Liu, Tao Tu, Paul Sajda

Simultaneous EEG-fMRI is a multi-modal neuroimaging technique that provides complementary spatial and temporal resolution. Challenging has been developing principled and interpretable approaches for fusing the modalities, specifically approaches enabling inference of latent source spaces representative of neural activity. In this paper, we address this inference problem within the framework of transcoding -- mapping from a specific encoding (modality) to a decoding (the latent source space) and then encoding the latent source space to the other modality. Specifically, we develop a symmetric method consisting of a cyclic convolutional transcoder that transcodes EEG to fMRI and vice versa. Without any prior knowledge of either the hemodynamic response function or lead field matrix, the complete data-driven method exploits the temporal and spatial relationships between the modalities and latent source spaces to learn these mappings. We quantify, for both the simulated and real EEG-fMRI data, how well the modalities can be transcoded from one to another as well as the source spaces that are recovered, all evaluated on unseen data. In addition to enabling a new way to symmetrically infer a latent source space, the method can also be seen as low-cost computational neuroimaging -- i.e. generating an 'expensive' fMRI BOLD image from 'low cost' EEG data.

Arunesh Mittal, Kai Yang, Paul Sajda et al.

Several approximate inference methods have been proposed for deep discrete latent variable models. However, non-parametric methods which have previously been successfully employed for classical sparse coding models have largely been unexplored in the context of deep models. We propose a non-parametric iterative algorithm for learning discrete latent representations in such deep models. Additionally, to learn scale invariant discrete features, we propose local data scaling variables. Lastly, to encourage sparsity in our representations, we propose a Beta-Bernoulli process prior on the latent factors. We evaluate our spare coding model coupled with different likelihood models. We evaluate our method across datasets with varying characteristics and compare our results to current amortized approximate inference methods.

Ankur Samanta, Akshayaa Magesh, Ayush Jain et al.

Contemporary reinforcement learning with verifiable reward methods post-train language models on multi-step reasoning by assigning a single outcome reward uniformly across all tokens in a trajectory. Such uniform assignment ignores which steps contributed to success or failure. Improving credit assignment can address this limitation by enabling targeted refinement of faulty reasoning steps, rather than updating entire trajectories uniformly. Resets are one such simple mechanism, enabling more precise credit assignment by returning to an intermediate state and resampling counterfactual continuations, so that outcome differences can be attributed to decisions made at that point. We propose two such methods: Random-Reset Policy Optimization (RRPO), where reset states are drawn randomly from reasoning steps, and Self-Reset Policy Optimization (SRPO), where the model self-localizes the erroneous step in an incorrect trajectory and resets there. We analyze these methods within the Conservative Policy Iteration (CPI) framework. Extending CPI with a credit-assignment oracle that targets improvable states yields provable improvements over random resets. Across models and reasoning benchmarks, SRPO consistently outperforms standard GRPO and RRPO by sampling multiple suffix continuations at a self-localized reset and learning from their rewards, using only the model itself with no external supervision.

Ankur Samanta, Akshayaa Magesh, Ayush Jain et al.

Self-correction in language models remains elusive. In this work, we explore whether language models can explicitly localize errors in incorrect reasoning, as a path toward building AI systems that can effectively correct themselves. We introduce a prompting method that structures reasoning as discrete, semantically coherent thought steps, and show that models are able to reliably localize errors within this structure, while failing to do so in conventional, unstructured chain-of-thought reasoning. Motivated by how the human brain monitors errors at discrete decision points and resamples alternatives, we introduce Iterative Correction Sampling of Thoughts (Thought-ICS), a self-correction framework. Thought-ICS iteratively prompts the model to generate reasoning one discrete and complete thought at a time--where each thought represents a deliberate decision by the model--creating natural boundaries for precise error localization. Upon verification, the model localizes the first erroneous step, and the system backtracks to generate alternative reasoning from the last correct point. When asked to correct reasoning verified as incorrect by an oracle, Thought-ICS achieves 20-40% self-correction lift. In a completely autonomous setting without external verification, it outperforms contemporary self-correction baselines.

Ziheng "Leo" Li, Xichen He, Mengyuan "Millie" Wu et al.

Despite steady progress, text entry in Extended Reality (XR) often remains slower and more effortful than typing on a physical keyboard or touchscreen. We explore a simple idea: use gaze to swipe through a virtual keyboard for the fast, low-effort where and a manual pinch held throughout the swipe for the when, extending and validating it through a series of user studies. We first show that a basic version including a low-latency decoder with spatiotemporal Dynamic Time Warping and fixation filtering outperforms selecting individual keys sequentially, either by finger tapping each or gazing at each while pinching. We then add mid-swipe prediction and in-gesture cancellation, improving words per minute (WPM) without hurting accuracy. We show that this approach is faster and more preferred than previous gaze-swipe approaches, finger tapping with prediction, or hand swiping with the same additions. Furthermore, a seven-day, 30-session study demonstrates sustained learning, with peak performance reaching 64.7 WPM.

Nikolas Papadopoulos, Shreenithi Navaneethan, Sheng Bai et al.

Preference learning methods, such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO), rely on pairwise human judgments, yet little is known about the cognitive processes underlying these judgments. We investigate whether eye-tracking can reveal preference formation during pairwise AI-generated image evaluation. Thirty participants completed 1,800 trials while their gaze was recorded. We replicated the gaze cascade effect, with gaze shifting toward chosen images approximately one second before the decision. Cascade dynamics were consistent across confidence levels. Gaze features predicted binary choice (68% accuracy), with chosen images receiving more dwell time, fixations, and revisits. Gaze transitions distinguished high-confidence from uncertain decisions (66% accuracy), with low-confidence trials showing more image switches per second. These results show that gaze patterns predict both choice and confidence in pairwise image evaluations, suggesting that eye-tracking provides implicit signals relevant to the quality of preference annotations.

Yinuo Qin, Richard T. Lee, Weijia Zhang et al.

In collaborative environments, a deep understanding of multi-human teaming dynamics is essential for optimizing performance. However, the relationship between individuals' behavioral and physiological markers and their combined influence on overall team performance remains poorly understood. To explore this, we designed a triadic human collaborative sensorimotor task in virtual reality (VR) and introduced a novel predictability metric to examine team dynamics and performance. Our findings reveal a strong connection between team performance and the predictability of a team member's future actions based on other team members' behavioral and physiological data. Contrary to conventional wisdom that high-performing teams are highly synchronized, our results suggest that physiological and behavioral synchronizations among team members have a limited correlation with team performance. These insights provide a new quantitative framework for understanding multi-human teaming, paving the way for deeper insights into team dynamics and performance.

Ankur Samanta, Akshayaa Magesh, Youliang Yu et al.

Language Models (LMs) are inconsistent reasoners, often generating contradictory responses to identical prompts. While inference-time methods can mitigate these inconsistencies, they fail to address the core problem: LMs struggle to reliably select reasoning pathways leading to consistent outcomes under exploratory sampling. To address this, we formalize self-consistency as an intrinsic property of well-aligned reasoning models and introduce Multi-Agent Consensus Alignment (MACA), a reinforcement learning framework that post-trains models to favor reasoning trajectories aligned with their internal consensus using majority/minority outcomes from multi-agent debate. These trajectories emerge from deliberative exchanges where agents ground reasoning in peer arguments, not just aggregation of independent attempts, creating richer consensus signals than single-round majority voting. MACA enables agents to teach themselves to be more decisive and concise, and better leverage peer insights in multi-agent settings without external supervision, driving substantial improvements across self-consistency (+27.6% on GSM8K), single-agent reasoning (+23.7% on MATH), sampling-based inference (+22.4% Pass@20 on MATH), and multi-agent ensemble decision-making (+42.7% on MathQA). These findings, coupled with strong generalization to unseen benchmarks (+16.3% on GPQA, +11.6% on CommonsenseQA), demonstrate robust self-alignment that more reliably unlocks latent reasoning potential of language models.

Sharath Koorathota, Arunesh Mittal, Richard P. Sloan et al.

Cognition in midlife is an important predictor of age-related mental decline and statistical models that predict cognitive performance can be useful for predicting decline. However, existing models struggle to capture complex relationships between physical, sociodemographic, psychological and mental health factors that effect cognition. Using data from an observational, cohort study, Midlife in the United States (MIDUS), we modeled a large number of variables to predict executive function and episodic memory measures. We used cross-sectional and longitudinal outcomes with varying sparsity, or amount of missing data. Deep neural network (DNN) models consistently ranked highest in all of the cognitive performance prediction tasks, as assessed with root mean squared error (RMSE) on out-of-sample data. RMSE differences between DNN and other model types were statistically significant (T(8) = -3.70; p < 0.05). The interaction effect between model type and sparsity was significant (F(9)=59.20; p < 0.01), indicating the success of DNNs can partly be attributed to their robustness and ability to model hierarchical relationships between health-related factors. Our findings underscore the potential of neural networks to model clinical datasets and allow better understanding of factors that lead to cognitive decline.

Arunesh Mittal, Scott Linderman, John Paisley et al.

We propose a hierarchical Bayesian recurrent state space model for modeling switching network connectivity in resting state fMRI data. Our model allows us to uncover shared network patterns across disease conditions. We evaluate our method on the ADNI2 dataset by inferring latent state patterns corresponding to altered neural circuits in individuals with Mild Cognitive Impairment (MCI). In addition to states shared across healthy and individuals with MCI, we discover latent states that are predominantly observed in individuals with MCI. Our model outperforms current state of the art deep learning method on ADNI2 dataset.

Arunesh Mittal, Paul Sajda, John Paisley

We propose a deep generative factor analysis model with beta process prior that can approximate complex non-factorial distributions over the latent codes. We outline a stochastic EM algorithm for scalable inference in a specific instantiation of this model and present some preliminary results.

Xueqing Liu, Linbi Hong, Paul Sajda

Simultaneous EEG-fMRI is a multi-modal neuroimaging technique that provides complementary spatial and temporal resolution for inferring a latent source space of neural activity. In this paper we address this inference problem within the framework of transcoding -- mapping from a specific encoding (modality) to a decoding (the latent source space) and then encoding the latent source space to the other modality. Specifically, we develop a symmetric method consisting of a cyclic convolutional transcoder that transcodes EEG to fMRI and vice versa. Without any prior knowledge of either the hemodynamic response function or lead field matrix, the method exploits the temporal and spatial relationships between the modalities and latent source spaces to learn these mappings. We show, for real EEG-fMRI data, how well the modalities can be transcoded from one to another as well as the source spaces that are recovered, all on unseen data. In addition to enabling a new way to symmetrically infer a latent source space, the method can also be seen as low-cost computational neuroimaging -- i.e. generating an 'expensive' fMRI BOLD image from 'low cost' EEG data.

Xueqing Liu, Paul Sajda

Many imaging technologies rely on tomographic reconstruction, which requires solving a multidimensional inverse problem given a finite number of projections. Backprojection is a popular class of algorithm for tomographic reconstruction, however it typically results in poor image reconstructions when the projection angles are sparse and/or if the sensors characteristics are not uniform. Several deep learning based algorithms have been developed to solve this inverse problem and reconstruct the image using a limited number of projections. However these algorithms typically require examples of the ground-truth (i.e. examples of reconstructed images) to yield good performance. In this paper, we introduce an unsupervised sparse-view backprojection algorithm, which does not require ground-truth. The algorithm consists of two modules in a generator-projector framework; a convolutional neural network and a spatial transformer network. We evaluated our algorithm using computed tomography (CT) images of the human chest. We show that our algorithm significantly out-performs filtered backprojection when the projection angles are very sparse, as well as when the sensor characteristics vary for different angles. Our approach has practical applications for medical imaging and other imaging modalities (e.g. radar) where sparse and/or non-uniform projections may be acquired due to time or sampling constraints.

Iretiayo Akinola, Zizhao Wang, Junyao Shi et al.

In reinforcement learning (RL), sparse rewards are a natural way to specify the task to be learned. However, most RL algorithms struggle to learn in this setting since the learning signal is mostly zeros. In contrast, humans are good at assessing and predicting the future consequences of actions and can serve as good reward/policy shapers to accelerate the robot learning process. Previous works have shown that the human brain generates an error-related signal, measurable using electroencephelography (EEG), when the human perceives the task being done erroneously. In this work, we propose a method that uses evaluative feedback obtained from human brain signals measured via scalp EEG to accelerate RL for robotic agents in sparse reward settings. As the robot learns the task, the EEG of a human observer watching the robot attempts is recorded and decoded into noisy error feedback signal. From this feedback, we use supervised learning to obtain a policy that subsequently augments the behavior policy and guides exploration in the early stages of RL. This bootstraps the RL learning process to enable learning from sparse reward. Using a robotic navigation task as a test bed, we show that our method achieves a stable obstacle-avoidance policy with high success rate, outperforming learning from sparse rewards only that struggles to achieve obstacle avoidance behavior or fails to advance to the goal.

Nicholas R. Waytowich, Vernon Lawhern, Javier O. Garcia et al.

Steady-State Visual Evoked Potentials (SSVEPs) are neural oscillations from the parietal and occipital regions of the brain that are evoked from flickering visual stimuli. SSVEPs are robust signals measurable in the electroencephalogram (EEG) and are commonly used in brain-computer interfaces (BCIs). However, methods for high-accuracy decoding of SSVEPs usually require hand-crafted approaches that leverage domain-specific knowledge of the stimulus signals, such as specific temporal frequencies in the visual stimuli and their relative spatial arrangement. When this knowledge is unavailable, such as when SSVEP signals are acquired asynchronously, such approaches tend to fail. In this paper, we show how a compact convolutional neural network (Compact-CNN), which only requires raw EEG signals for automatic feature extraction, can be used to decode signals from a 12-class SSVEP dataset without the need for any domain-specific knowledge or calibration data. We report across subject mean accuracy of approximately 80% (chance being 8.3%) and show this is substantially better than current state-of-the-art hand-crafted approaches using canonical correlation analysis (CCA) and Combined-CCA. Furthermore, we analyze our Compact-CNN to examine the underlying feature representation, discovering that the deep learner extracts additional phase and amplitude related features associated with the structure of the dataset. We discuss how our Compact-CNN shows promise for BCI applications that allow users to freely gaze/attend to any stimulus at any time (e.g., asynchronous BCI) as well as provides a method for analyzing SSVEP signals in a way that might augment our understanding about the basic processing in the visual cortex.

Victor Shih, David C Jangraw, Paul Sajda et al.

Reinforcement Learning AI commonly uses reward/penalty signals that are objective and explicit in an environment -- e.g. game score, completion time, etc. -- in order to learn the optimal strategy for task performance. However, Human-AI interaction for such AI agents should include additional reinforcement that is implicit and subjective -- e.g. human preferences for certain AI behavior -- in order to adapt the AI behavior to idiosyncratic human preferences. Such adaptations would mirror naturally occurring processes that increase trust and comfort during social interactions. Here, we show how a hybrid brain-computer-interface (hBCI), which detects an individual's level of interest in objects/events in a virtual environment, can be used to adapt the behavior of a Deep Reinforcement Learning AI agent that is controlling a virtual autonomous vehicle. Specifically, we show that the AI learns a driving strategy that maintains a safe distance from a lead vehicle, and most novelly, preferentially slows the vehicle when the human passengers of the vehicle encounter objects of interest. This adaptation affords an additional 20\% viewing time for subjectively interesting objects. This is the first demonstration of how an hBCI can be used to provide implicit reinforcement to an AI agent in a way that incorporates user preferences into the control system.

Josef Faller, Linbi Hong, Jennifer Cummings et al.

Simultaneously recorded electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) can be used to non-invasively measure the spatiotemporal dynamics of the human brain. One challenge is dealing with the artifacts that each modality introduces into the other when the two are recorded concurrently, for example the ballistocardiogram (BCG). We conducted a preliminary comparison of three different MR compatible EEG recording systems and assessed their performance in terms of single-trial classification of the EEG when simultaneously collecting fMRI. We found tradeoffs across all three systems, for example varied ease of setup and improved classification accuracy with reference electrodes (REF) but not for pulse artifact subtraction (PAS) or reference layer adaptive filtering (RLAF).

Bryan R. Conroy, Jennifer M. Walz, Brian Cheung et al.

We present an efficient algorithm for simultaneously training sparse generalized linear models across many related problems, which may arise from bootstrapping, cross-validation and nonparametric permutation testing. Our approach leverages the redundancies across problems to obtain significant computational improvements relative to solving the problems sequentially by a conventional algorithm. We demonstrate our fast simultaneous training of generalized linear models (FaSTGLZ) algorithm on a number of real-world datasets, and we run otherwise computationally intensive bootstrapping and permutation test analyses that are typically necessary for obtaining statistically rigorous classification results and meaningful interpretation. Code is freely available at http://liinc.bme.columbia.edu/fastglz.