📌 TOPINDIATOURS Hot ai: Which Agent Causes Task Failures and When?Researchers from

Share My Research is Synced’s column that welcomes scholars to share their own research breakthroughs with over 1.5M global AI enthusiasts. Beyond technological advances, Share My Research also calls for interesting stories behind the research and exciting research ideas. Contact us: chain.zhang@jiqizhixin.com

Meet the authors
Institutions: Penn State University, Duke University, Google DeepMind, University of Washington, Meta, Nanyang Technological University, and Oregon State University. The co-first authors are Shaokun Zhang of Penn State University and Ming Yin of Duke University.

In recent years, LLM Multi-Agent systems have garnered widespread attention for their collaborative approach to solving complex problems. However, it’s a common scenario for these systems to fail at a task despite a flurry of activity. This leaves developers with a critical question: which agent, at what point, was responsible for the failure? Sifting through vast interaction logs to pinpoint the root cause feels like finding a needle in a haystack—a time-consuming and labor-intensive effort.
 
This is a familiar frustration for developers. In increasingly complex Multi-Agent systems, failures are not only common but also incredibly difficult to diagnose due to the autonomous nature of agent collaboration and long information chains. Without a way to quickly identify the source of a failure, system iteration and optimization grind to a halt.
 
To address this challenge, researchers from Penn State University and Duke University, in collaboration with institutions including Google DeepMind, have introduced the novel research problem of “Automated Failure Attribution.” They have constructed the first benchmark dataset for this task, Who&When, and have developed and evaluated several automated attribution methods. This work not only highlights the complexity of the task but also paves a new path toward enhancing the reliability of LLM Multi-Agent systems.

The paper has been accepted as a Spotlight presentation at the top-tier machine learning conference, ICML 2025, and the code and dataset are now fully open-source.

Paper：https://arxiv.org/pdf/2505.00212
Code：https://github.com/mingyin1/Agents_Failure_Attribution
Dataset：https://huggingface.co/datasets/Kevin355/Who_and_When
 
 
Research Background and Challenges
LLM-driven Multi-Agent systems have demonstrated immense potential across many domains. However, these systems are fragile; errors by a single agent, misunderstandings between agents, or mistakes in information transmission can lead to the failure of the entire task.

Currently, when a system fails, developers are often left with manual and inefficient methods for debugging:
Manual Log Archaeology : Developers must manually review lengthy interaction logs to find the source of the problem.
Reliance on Expertise : The debugging process is highly dependent on the developer’s deep understanding of the system and the task at hand.
 
This “needle in a haystack” approach to debugging is not only inefficient but also severely hinders rapid system iteration and the improvement of system reliability. There is an urgent need for an automated, systematic method to pinpoint the cause of failures, effectively bridging the gap between “evaluation results” and “system improvement.”

Core Contributions
This paper makes several groundbreaking contributions to address the challenges above:
1. Defining a New Problem: The paper is the first to formalize “automated failure attribution” as a specific research task. This task is defined by identifying the failure-responsible agent and the decisive error step that led to the task’s failure.
2. Constructing the First Benchmark Dataset: Who&When : This dataset includes a wide range of failure logs collected from 127 LLM Multi-Agent systems, which were either algorithmically generated or hand-crafted by experts to ensure realism and diversity. Each failure log is accompanied by fine-grained human annotations for:
Who: The agent responsible for the failure.
When: The specific interaction step where the decisive error occurred.
Why: A natural language explanation of the cause of the failure.

3. Exploring Initial “Automated Attribution” Methods : Using the Who&When dataset, the paper designs and assesses three distinct methods for automated failure attribution:
– All-at-Once: This method provides the LLM with the user query and the complete failure log, asking it to identify the responsible agent and the decisive error step in a single pass. While cost-effective, it may struggle to pinpoint precise errors in long contexts.
– Step-by-Step: This approach mimics manual debugging by having the LLM review the interaction log sequentially, making a judgment at each step until the error is found. It is more precise at locating the error step but incurs higher costs and risks accumulating errors.
– Binary Search: A compromise between the first two methods, this strategy repeatedly divides the log in half, using the LLM to determine which segment contains the error. It then recursively searches the identified segment, offering a balance of cost and performance.

Experimental Results and Key Findings 
Experiments were conducted in two settings: one where the LLM knows the ground truth answer to the problem the Multi-Agent system is trying to solve (With Ground Truth) and one where it does not (Without Ground Truth). The primary model used was GPT-4o, though other models were also tested. The systematic evaluation of these methods on the Who&When dataset yielded several important insights:
– A Long Way to Go: Current methods are far from perfect. Even the best-performing single method achieved an accuracy of only about 53.5% in identifying the responsible agent and a mere 14.2% in pinpointing the exact error step. Some methods performed even worse than random guessing, underscoring the difficulty of the task.
– No “All-in-One” Solution: Different methods excel at different aspects of the problem. The All-at-Once method is better at identifying “Who,” while the Step-by-Step method is more effective at determining “When.” The Binary Search method provides a middle-ground performance.
 

– Hybrid Approaches Show Promise but at a High Cost: The researchers found that combining different methods, such as using the All-at-Once approach to identify a potential agent and then applying the Step-by-Step method to find the error, can improve overall performance. However, this comes with a significant increase in computational cost.

– State-of-the-Art Models Struggle: Surprisingly, even the most advanced reasoning models, like OpenAI o1 and DeepSeek R1, find this task challenging.- This h…

Konten dipersingkat otomatis.

🔗 Sumber: syncedreview.com

📌 TOPINDIATOURS Update ai: Chinese scientists build mini womb on a chip to study h

Every human life starts with a risky and delicate event—a microscopic embryo finding a place to settle inside the uterus. If this process fails, pregnancy never begins, no matter how healthy the embryo looks. 

For decades, scientists have struggled to study this moment in humans because of ethical restrictions and limited access to early pregnancy tissue. 

Now, a team of researchers in China has crossed this barrier by producing a miniature womb on a chip that faithfully recreates how human embryos attach to and burrow into the uterine lining.

“This system successfully recapitulates key events of human implantation and early post-implantation development,” the researchers note. 

This advancement opens a new way to understanding infertility, improving IVF success, and testing treatments in ways that weren’t possible before.

Implantation has been almost impossible to study

Embryo implantation happens just days after fertilisation, when a five- to six-day-old embryo reaches the uterus. At this stage, the embryo must complete three tightly controlled steps. It first brushes against the uterine lining, then locks onto it, and finally pushes its way inside to establish a connection that will support pregnancy.

Studying these steps directly in humans is extremely difficult. Ethical rules limit experiments on natural embryos, and early pregnancy tissue is usually available only from rare medical procedures such as hysterectomies. 

Existing lab models, including flat cell cultures and endometrial organoids, capture only fragments of the process and miss the complex three-dimensional interaction between embryo and uterus. 

As a result, many cases of implantation failure, especially in IVF, remain poorly understood. To address this gap, researchers from the Chinese Academy of Sciences designed a three-dimensional model of the human endometrium, the tissue that lines the uterus. 

Creating the uterus lining on a chip

They began by embedding human endometrial cells into gel-like layers, allowing the cells to grow and organise themselves into a structure that closely resembles real uterine tissue. This engineered tissue, known as an endometrioid, was placed inside a microfluidic chip—a small device that can control the movement of fluids and nutrients. 

The chip environment allowed the tissue to behave more as it does inside the body, rather than in a traditional petri dish. More importantly, the endometrial cells used to build the model could be obtained from a single biopsy, and the system was also compatible with cells collected non-invasively from menstrual blood.

Once the artificial uterine lining was ready, the team introduced two types of embryos into the chip. One was a real human blastocysts, which contain about 100 to 200 rapidly dividing cells. 

The other was blastoids, lab-made structures created from stem cells that closely mimic natural blastocysts and can be produced in large numbers with consistent genetic properties. Inside the chip, both blastocysts and blastoids went through the full implantation sequence. 

They made initial contact with the uterine surface, formed stable attachments using molecular signals, and then actively invaded the tissue, embedding themselves just as they would in early pregnancy. This level of detail had not been achieved with earlier two-dimensional models. 

“Our in-chip 3D endometrioid-based implantation model offers a streamlined platform that captures all major stages of implantation – apposition, attachment, and invasion – as well as early post-implantation development,” the study authors said.

When the researchers built womb chips using cells from women diagnosed with recurrent implantation failure—defined as repeated IVF failures—the embryos showed a much lower ability to implant. This mirrored real-world clinical outcomes, confirming that the model can capture patient-specific differences. 

The team then used the platform to screen over 1,000 FDA-approved drugs and identified compounds that improved implantation performance, demonstrating the chip’s value as a drug-testing tool.

A way of improving infertility treatment

With infertility affecting roughly one in six adults worldwide, understanding why embryos fail to implant is a major medical priority. This womb-on-a-chip approach offers a way to study implantation safely, ethically, and in a highly controlled setting. 

It could help doctors identify why IVF fails in certain patients and tailor treatments based on individual uterine responses rather than trial and error. 

However, this approach still has limitations. It does not yet include blood vessels or immune cells, both of which play crucial roles in reshaping the uterus and supporting a growing placenta. 

Hopefully, the researchers will incorporate these missing elements in future studies to make the model even closer to real human biology.

The study is published in the journal Cell.

🔗 Sumber: interestingengineering.com