Jingya Chen

Vidhisha Balachandran, Jingya Chen, Neel Joshi et al. · cmu

Rigorous and reproducible evaluation is critical for assessing the state of the art and for guiding scientific advances in Artificial Intelligence. Evaluation is challenging in practice due to several reasons, including benchmark saturation, lack of transparency in methods used for measurement, development challenges in extracting measurements for generative tasks, and, more generally, the extensive number of capabilities required for a well-rounded comparison across models. We make three contributions to alleviate the above challenges. First, we present Eureka, an open-source framework for standardizing evaluations of large foundation models beyond single-score reporting and rankings. Second, we introduce Eureka-Bench as an extensible collection of benchmarks testing capabilities that (i) are still challenging for state-of-the-art models and (ii) represent fundamental but overlooked language and multimodal capabilities. The inherent space for improvement in non-saturated benchmarks enables us to discover meaningful differences between models at a capability level. Third, using Eureka, we conduct an analysis of 12 state-of-the-art models, providing in-depth insights into failure understanding and model comparison, which can be leveraged to plan targeted improvements. In contrast to recent trends in reports and leaderboards showing absolute rankings and claims for one model or another to be the best, our analysis shows that there is no such best model. Different models have different strengths, but there are models that appear more often than others as best performers for some capabilities. Despite the recent improvements, current models still struggle with several fundamental capabilities including detailed image understanding, benefiting from multimodal input when available rather than fully relying on language, factuality and grounding for information retrieval, and over refusals.

Victor Dibia, Jingya Chen, Gagan Bansal et al.

Multi-agent systems, where multiple agents (generative AI models + tools) collaborate, are emerging as an effective pattern for solving long-running, complex tasks in numerous domains. However, specifying their parameters (such as models, tools, and orchestration mechanisms etc,.) and debugging them remains challenging for most developers. To address this challenge, we present AUTOGEN STUDIO, a no-code developer tool for rapidly prototyping, debugging, and evaluating multi-agent workflows built upon the AUTOGEN framework. AUTOGEN STUDIO offers a web interface and a Python API for representing LLM-enabled agents using a declarative (JSON-based) specification. It provides an intuitive drag-and-drop UI for agent workflow specification, interactive evaluation and debugging of workflows, and a gallery of reusable agent components. We highlight four design principles for no-code multi-agent developer tools and contribute an open-source implementation at https://github.com/microsoft/autogen/tree/main/samples/apps/autogen-studio

Madeleine Grunde-McLaughlin, Hussein Mozannar, Maya Murad et al. · microsoft-research

To enable human oversight, agentic AI systems often provide a trace of reasoning and action steps. Designing traces to have an informative, but not overwhelming, level of detail remains a critical challenge. In three user studies on a Computer User Agent, we investigate the utility of basic action traces for verification, explore three alternatives via design probes, and test a novel interface's impact on error finding in question-answering tasks. As expected, we find that current practices are cumbersome, limiting their efficacy. Conversely, our proposed design reduced the time participants spent finding errors. However, although participants reported higher levels of confidence in their decisions, their final accuracy was not meaningfully improved. To this end, our study surfaces challenges for human verification of agentic systems, including managing built-in assumptions, users' subjective and changing correctness criteria, and the shortcomings, yet importance, of communicating the agent's process.

Bingxue Wang, Liwen Zou, Jun Chen et al.

The presence of tertiary lymphoid structures (TLSs) on pancreatic pathological images is an important prognostic indicator of pancreatic tumors. Therefore, TLSs detection on pancreatic pathological images plays a crucial role in diagnosis and treatment for patients with pancreatic tumors. However, fully supervised detection algorithms based on deep learning usually require a large number of manual annotations, which is time-consuming and labor-intensive. In this paper, we aim to detect the TLSs in a manner of few-shot learning by proposing a weakly supervised segmentation network. We firstly obtain the lymphocyte density maps by combining a pretrained model for nuclei segmentation and a domain adversarial network for lymphocyte nuclei recognition. Then, we establish a cross-scale attention guidance mechanism by jointly learning the coarse-scale features from the original histopathology images and fine-scale features from our designed lymphocyte density attention. A noise-sensitive constraint is introduced by an embedding signed distance function loss in the training procedure to reduce tiny prediction errors. Experimental results on two collected datasets demonstrate that our proposed method significantly outperforms the state-of-the-art segmentation-based algorithms in terms of TLSs detection accuracy. Additionally, we apply our method to study the congruent relationship between the density of TLSs and peripancreatic vascular invasion and obtain some clinically statistical results.

Sam Yu-Te Lee, Jingya Chen, Albert Calzaretto et al.

Enterprise chatbots show promise in supporting knowledge workers in information synthesis tasks by retrieving context from large, heterogeneous databases before generating answers. However, when the retrieved context misaligns with user intentions, the chatbot often produces "irrelevantly right" responses that provide little value. In this work, we introduce VizCopilot, a prototype that incorporates visualization techniques to actively involve end-users in context alignment. By combining topic modeling with document visualization, VizCopilot enables human oversight and modification of retrieved context while keeping cognitive overhead manageable. We used VizCopilot as a design probe in a Research-through-Design study to evaluate the role of visualization in context alignment and to surface future design opportunities. Our findings show that visualization not only helps users detect and correct misaligned context but also encourages them to adapt their prompting strategies, enabling the system to retrieve more relevant context from the outset. At the same time, the study reveals limitations in verification support regarding close-reading and trust in AI summaries. We outline future directions for visualization-enhanced chatbots, focusing on personalization, proactivity, and sustainable human-AI collaboration.

Hussein Mozannar, Gagan Bansal, Cheng Tan et al. · microsoft-research

AI agents powered by large language models are increasingly capable of autonomously completing complex, multi-step tasks using external tools. Yet, they still fall short of human-level performance in most domains including computer use, software development, and research. Their growing autonomy and ability to interact with the outside world, also introduces safety and security risks including potentially misaligned actions and adversarial manipulation. We argue that human-in-the-loop agentic systems offer a promising path forward, combining human oversight and control with AI efficiency to unlock productivity from imperfect systems. We introduce Magentic-UI, an open-source web interface for developing and studying human-agent interaction. Built on a flexible multi-agent architecture, Magentic-UI supports web browsing, code execution, and file manipulation, and can be extended with diverse tools via Model Context Protocol (MCP). Moreover, Magentic-UI presents six interaction mechanisms for enabling effective, low-cost human involvement: co-planning, co-tasking, multi-tasking, action guards, and long-term memory. We evaluate Magentic-UI across four dimensions: autonomous task completion on agentic benchmarks, simulated user testing of its interaction capabilities, qualitative studies with real users, and targeted safety assessments. Our findings highlight Magentic-UI's potential to advance safe and efficient human-agent collaboration.

Kyle O'Brien, David Majercak, Xavier Fernandes et al.

Responsible deployment of language models requires mechanisms for refusing unsafe prompts while preserving model performance. While most approaches modify model weights through additional training, we explore an alternative: steering model activations at inference time via amplifying sparse autoencoder (SAE) features that mediate refusal. This work uncovers a fundamental tension between SAE steering-based safety improvements and general model capabilities. While feature steering successfully improves robustness against both single-turn and challenging multi-turn jailbreak attempts, we discover that this comes at a previously underexplored cost -- systematic degradation of performance across multiple benchmark tasks, even on safe inputs with no apparent connection to refusal behavior. This suggests that features mediating refusal may be more deeply entangled with general language model capabilities than previously understood. Our findings reveal important open questions about the nature of safety-relevant features in language models and the feasibility of isolating them for targeted intervention. While SAE-based steering shows promise as a flexible approach to enhancing language model safety, our results highlight the critical need to understand and address the mechanisms behind these capability tradeoffs before such techniques can be practically deployed.

Vidhisha Balachandran, Jingya Chen, Lingjiao Chen et al. · cmu

Inference-time scaling can enhance the reasoning capabilities of large language models (LLMs) on complex problems that benefit from step-by-step problem solving. Although lengthening generated scratchpads has proven effective for mathematical tasks, the broader impact of this approach on other tasks remains less clear. In this work, we investigate the benefits and limitations of scaling methods across nine state-of-the-art models and eight challenging tasks, including math and STEM reasoning, calendar planning, NP-hard problems, navigation, and spatial reasoning. We compare conventional models (e.g., GPT-4o) with models fine-tuned for inference-time scaling (e.g., o1) through evaluation protocols that involve repeated model calls, either independently or sequentially with feedback. These evaluations approximate lower and upper performance bounds and potential for future performance improvements for each model, whether through enhanced training or multi-model inference systems. Our extensive empirical analysis reveals that the advantages of inference-time scaling vary across tasks and diminish as problem complexity increases. In addition, simply using more tokens does not necessarily translate to higher accuracy in these challenging regimes. Results from multiple independent runs with conventional models using perfect verifiers show that, for some tasks, these models can achieve performance close to the average performance of today's most advanced reasoning models. However, for other tasks, a significant performance gap remains, even in very high scaling regimes. Encouragingly, all models demonstrate significant gains when inference is further scaled with perfect verifiers or strong feedback, suggesting ample potential for future improvements.

Toby Jia-Jun Li, Jingya Chen, Brandon Canfield et al.

An important concern in end user development (EUD) is accidentally embedding personal information in program artifacts when sharing them. This issue is particularly important in GUI-based programming-by-demonstration (PBD) systems due to the lack of direct developer control of script contents. Prior studies reported that these privacy concerns were the main barrier to script sharing in EUD. We present a new approach that can identify and obfuscate the potential personal information in GUI-based PBD scripts based on the uniqueness of information entries with respect to the corresponding app GUI context. Compared with the prior approaches, ours supports broader types of personal information beyond explicitly pre-specified ones, requires minimal user effort, addresses the threat of re-identification attacks, and can work with third-party apps from any task domain. Our approach also recovers obfuscated fields locally on the script consumer's side to preserve the shared scripts' transparency, readability, robustness, and generalizability. Our evaluation shows that our approach (1) accurately identifies the potential personal information in scripts across different apps in diverse task domains; (2) allows end-user developers to feel comfortable sharing their own scripts; and (3) enables script consumers to understand the operation of shared scripts despite the obfuscated fields.

Toby Jia-Jun Li, Jingya Chen, Tom M. Mitchell et al.

We argue that a key challenge in enabling usable and useful interactive task learning for intelligent agents is to facilitate effective Human-AI collaboration. We reflect on our past 5 years of efforts on designing, developing and studying the SUGILITE system, discuss the issues on incorporating recent advances in AI with HCI principles in mixed-initiative interactions and multi-modal interactions, and summarize the lessons we learned. Lastly, we identify several challenges and opportunities, and describe our ongoing work