Rishabh Agrawal

Amazon AGI, Aaron Langford, Aayush Shah et al. · amazon-science

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

Rishabh Agrawal, Jacob Fein-Ashley, Paria Rashidinejad

Reasoning models have advanced rapidly, but the dominant reinforcement learning from verifiable rewards (RLVR) recipe remains surprisingly narrow: sample many responses and reward each with a single bit indicating whether the final answer is correct. Yet many settings provide rich feedback, including execution traces, tool outputs, expert corrections, and model self-evaluations. We study how to use such feedback through a distributional variant of the classic imitation learning algorithm DAgger, where the learner has local access to an expert distribution on states visited by the current policy. This yields a simple forward cross-entropy objective that admits a blackbox expert and whose sequence-level gradient {conduct rich credit assignment by propagating} future expert-student disagreement back to earlier decisions. We show that prior RL with self-distillation objectives based on reverse KL or Jensen-Shannon fail to guarantee monotonic policy improvement: even when the expert has higher reward, their updates may increase probability on worse actions. In contrast, we show that forward cross-entropy admits monotonic policy improvement and enjoys guarantees on regret. We further show that our objective optimizes a lower bound on teacher-weighted likelihood of success, leading to improved Pass@N. Empirically, our approach, DistIL, improves over RLVR and RL with self-distillation baselines across a variety of domains: scientific reasoning, coding, and solving hard mathematical problems.

Rishabh Agrawal, Nirav Ajmeri, Munindar P. Singh

Norms help regulate a society. Norms may be explicit (represented in structured form) or implicit. We address the emergence of explicit norms by developing agents who provide and reason about explanations for norm violations in deciding sanctions and identifying alternative norms. These agents use a genetic algorithm to produce norms and reinforcement learning to learn the values of these norms. We find that applying explanations leads to norms that provide better cohesion and goal satisfaction for the agents. Our results are stable for societies with differing attitudes of generosity.

Rishabh Agrawal, Nathan Dahlin, Rahul Jain et al.

Imitation Learning (IL) is an important paradigm within the broader reinforcement learning (RL) methodology. Unlike most of RL, it does not assume availability of reward-feedback. Reward inference and shaping are known to be difficult and error-prone methods particularly when the demonstration data comes from human experts. Classical methods such as behavioral cloning and inverse reinforcement learning are highly sensitive to estimation errors, a problem that is particularly acute in continuous state space problems. Meanwhile, state-of-the-art IL algorithms convert behavioral policy learning problems into distribution-matching problems which often require additional online interaction data to be effective. In this paper, we consider the problem of imitation learning in continuous state space environments based solely on observed behavior, without access to transition dynamics information, reward structure, or, most importantly, any additional interactions with the environment. Our approach is based on the Markov balance equation and introduces a novel conditional kernel density estimation-based imitation learning framework. It involves estimating the environment's transition dynamics using conditional kernel density estimators and seeks to satisfy the probabilistic balance equations for the environment. We establish that our estimators satisfy basic asymptotic consistency requirements. Through a series of numerical experiments on continuous state benchmark environments, we show consistently superior empirical performance over many state-of-the-art IL algorithms.

Rishabh Agrawal, Rahul Jain, Ashutosh Nayyar

Behavior Foundation Models (BFMs) enable scalable imitation learning (IL) by pretraining task-agnostic representations that can be rapidly adapted to new tasks. However, existing BFMs assume fixed environment dynamics, limiting their robustness under real-world shifts such as changes in friction, actuation, or sensor noise. We address this by formulating BFM task-inference as a robust minimax optimization problem, enabling adaptation to worst-case dynamics perturbations without modifying pretraining. To the best of our knowledge, this is the first BFM-based framework that achieves robustness to dynamics shifts while relying solely on offline data from a single nominal environment. Our approach significantly outperforms standard BFM and robust offline IL baselines under dynamics shifts. These results demonstrate that robust policy can be achieved entirely at task-inference time, improving the practicality of BFMs in dynamic settings.

Rishabh Agrawal, Nathan Dahlin, Rahul Jain et al.

Imitation learning (IL) is notably effective for robotic tasks where directly programming behaviors or defining optimal control costs is challenging. In this work, we address a scenario where the imitator relies solely on observed behavior and cannot make environmental interactions during learning. It does not have additional supplementary datasets beyond the expert's dataset nor any information about the transition dynamics. Unlike state-of-the-art (SOTA) IL methods, this approach tackles the limitations of conventional IL by operating in a more constrained and realistic setting. Our method uses the Markov balance equation and introduces a novel conditional density estimation-based imitation learning framework. It employs conditional normalizing flows for transition dynamics estimation and aims at satisfying a balance equation for the environment. Through a series of numerical experiments on Classic Control and MuJoCo environments, we demonstrate consistently superior empirical performance compared to many SOTA IL algorithms.

Xiao Ling, Rishabh Agrawal, Tim Menzies

Improved test case prioritization means that software developers can detect and fix more software faults sooner than usual. But is there one "best" prioritization algorithm? Or do different kinds of projects deserve special kinds of prioritization? To answer these questions, this paper applies nine prioritization schemes to 31 projects that range from (a) highly rated open-source Github projects to (b) computational science software to (c) a closed-source project. We find that prioritization approaches that work best for open-source projects can work worst for the closed-source project (and vice versa). From these experiments, we conclude that (a) it is ill-advised to always apply one prioritization scheme to all projects since (b) prioritization requires tuning to different project types.

Tianpei Xia, Wei Fu, Rui Shu et al.

Software developed on public platform is a source of data that can be used to make predictions about those projects. While the individual developing activity may be random and hard to predict, the developing behavior on project level can be predicted with good accuracy when large groups of developers work together on software projects. To demonstrate this, we use 64,181 months of data from 1,159 GitHub projects to make various predictions about the recent status of those projects (as of April 2020). We find that traditional estimation algorithms make many mistakes. Algorithms like $k$-nearest neighbors (KNN), support vector regression (SVR), random forest (RFT), linear regression (LNR), and regression trees (CART) have high error rates. But that error rate can be greatly reduced using hyperparameter optimization. To the best of our knowledge, this is the largest study yet conducted, using recent data for predicting multiple health indicators of open-source projects.

Aja Khanal, Kaushik T. Ranade, Rishabh Agrawal et al.

Generating high-quality structured data such as JSON records, remains a fundamental challenge for large language models (LLMs), particularly when semantic richness must coexist with strict schema adherence. While autoregressive LLMs offer strong structural consistency, they often struggle with semantic variation and output diversity. In contrast, diffusion language models (DLMs) introduce powerful mechanisms for semantic richness and bidirectional decoding, yet lack the inductive biases needed for reliable structure preservation. We present Agents of Diffusion (AoD), a novel framework that unifies the generative flexibility of DLMs with the reasoning capabilities of autoregressive models through language-mediated reinforcement learning. AoD frames structured text generation as a multi-agent alignment process, where a prompt optimization agent collaborates with a judge agent to iteratively guide a DLM using natural language feedback. This approach enables controllable, schema-consistent generation without modifying model parameters or relying on handcrafted constraints. AoD advances the state of controllable generation by demonstrating that diffusion models, when supervised by cooperative agents, can achieve both high semantic novelty and structural fidelity. Across multiple structured data benchmarks, AoD consistently outperforms diffusion and autoregressive baselines, establishing a new path forward for structure-aware, diversity-enhanced text synthesis.

Rishabh Agrawal, Yusuf Alvi, Rahul Jain et al.

Imitation Learning (IL) has proven highly effective for robotic and control tasks where manually designing reward functions or explicit controllers is infeasible. However, standard IL methods implicitly assume that the environment dynamics remain fixed between training and deployment. In practice, this assumption rarely holds where modeling inaccuracies, real-world parameter variations, and adversarial perturbations can all induce shifts in transition dynamics, leading to severe performance degradation. We address this challenge through Balance Equation-based Distributionally Robust Offline Imitation Learning, a framework that learns robust policies solely from expert demonstrations collected under nominal dynamics, without requiring further environment interaction. We formulate the problem as a distributionally robust optimization over an uncertainty set of transition models, seeking a policy that minimizes the imitation loss under the worst-case transition distribution. Importantly, we show that this robust objective can be reformulated entirely in terms of the nominal data distribution, enabling tractable offline learning. Empirical evaluations on continuous-control benchmarks demonstrate that our approach achieves superior robustness and generalization compared to state-of-the-art offline IL baselines, particularly under perturbed or shifted environments.

Rishabh Agrawal, Himanshu Kumar

The rapid progress in large language models (LLMs) has paved the way for novel approaches in knowledge-intensive tasks. Among these, Cache-Augmented Generation (CAG) has emerged as a promising alternative to Retrieval-Augmented Generation (RAG). CAG minimizes retrieval latency and simplifies system design by preloading knowledge into the model's context. However, challenges persist in scaling CAG to accommodate large and dynamic knowledge bases effectively. This paper introduces Adaptive Contextual Compression (ACC), an innovative technique designed to dynamically compress and manage context inputs, enabling efficient utilization of the extended memory capabilities of modern LLMs. To further address the limitations of standalone CAG, we propose a Hybrid CAG-RAG Framework, which integrates selective retrieval to augment preloaded contexts in scenarios requiring additional information. Comprehensive evaluations on diverse datasets highlight the proposed methods' ability to enhance scalability, optimize efficiency, and improve multi-hop reasoning performance, offering practical solutions for real-world knowledge integration challenges.

Rishabh Agrawal

Few-shot learning (FSL) enables machine learning models to generalize effectively with minimal labeled data, making it crucial for data-scarce domains such as healthcare, robotics, and natural language processing. Despite its potential, FSL faces challenges including sensitivity to initialization, difficulty in adapting to diverse domains, and vulnerability to noisy datasets. To address these issues, this paper introduces Adaptive Few-Shot Learning (AFSL), a framework that integrates advancements in meta-learning, domain alignment, noise resilience, and multi-modal integration. AFSL consists of four key modules: a Dynamic Stability Module for performance consistency, a Contextual Domain Alignment Module for domain adaptation, a Noise-Adaptive Resilience Module for handling noisy data, and a Multi-Modal Fusion Module for integrating diverse modalities. This work also explores strategies such as task-aware data augmentation, semi-supervised learning, and explainable AI techniques to enhance the applicability and robustness of FSL. AFSL provides scalable, reliable, and impactful solutions for real-world, high-stakes domains.

Amritanshu Agrawal, Xueqi Yang, Rishabh Agrawal et al.

How to make software analytics simpler and faster? One method is to match the complexity of analysis to the intrinsic complexity of the data being explored. For example, hyperparameter optimizers find the control settings for data miners that improve for improving the predictions generated via software analytics. Sometimes, very fast hyperparameter optimization can be achieved by just DODGE-ing away from things tried before. But when is it wise to use DODGE and when must we use more complex (and much slower) optimizers? To answer this, we applied hyperparameter optimization to 120 SE data sets that explored bad smell detection, predicting Github ssue close time, bug report analysis, defect prediction, and dozens of other non-SE problems. We find that DODGE works best for data sets with low "intrinsic dimensionality" (D = 3) and very poorly for higher-dimensional data (D over 8). Nearly all the SE data seen here was intrinsically low-dimensional, indicating that DODGE is applicable for many SE analytics tasks.

Huy Tu, Rishabh Agrawal, Tim Menzies

How should software engineering be adapted for Computational Science (CS)? If we understood that, then we could better support software sustainability, verifiability, reproducibility, comprehension, and usability for CS community. For example, improving the maintainability of the CS code could lead to: (a) faster adaptation of scientific project simulations to new and efficient hardware (multi-core and heterogeneous systems); (b) better support for larger teams to co-ordinate (through integration with interdisciplinary teams); and (c) an extended capability to model complex phenomena. In order to better understand computational science, this paper uses quantitative evidence (from 59 CS projects in Github) to check 13 published beliefs about CS. These beliefs reflect on (a) the nature of scientific challenges; (b) the implications of limitations of computer hardware; and (c) the cultural environment of scientific software development. What we found was, using this new data from Github, only a minority of those older beliefs can be endorsed. More than half of the pre-existing beliefs are dubious, which leads us to conclude that the nature of CS software development is changing. Further, going forward, this has implications for (1) what kinds of tools we would propose to better support computational science and (2) research directions for both communities.

Amritanshu Agrawal, Xueqi Yang, Rishabh Agrawal et al.

How can we make software analytics simpler and faster? One method is to match the complexity of analysis to the intrinsic complexity of the data being explored. For example, hyperparameter optimizers find the control settings for data miners that improve the predictions generated via software analytics. Sometimes, very fast hyperparameter optimization can be achieved by "DODGE-ing"; i.e. simply steering way from settings that lead to similar conclusions. But when is it wise to use that simple approach and when must we use more complex (and much slower) optimizers?} To answer this, we applied hyperparameter optimization to 120 SE data sets that explored bad smell detection, predicting Github issue close time, bug report analysis, defect prediction, and dozens of other non-SE problems. We find that the simple DODGE works best for data sets with low "intrinsic dimensionality" (u ~ 3) and very poorly for higher-dimensional data (u > 8). Nearly all the SE data seen here was intrinsically low-dimensional, indicating that DODGE is applicable for many SE analytics tasks.