MAROKO133 Hot ai: Which Agent Causes Task Failures and When?Researchers from PSU and Duke

📌 MAROKO133 Update 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: [email protected]

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

📌 MAROKO133 Update ai: Gemini 3 Flash arrives with reduced costs and latency — a p

Enterprises can now harness the power of a large language model that's near that of the state-of-the-art Google’s Gemini 3 Pro, but at a fraction of the cost and with increased speed, thanks to the newly released Gemini 3 Flash.

The model joins the flagship Gemini 3 Pro, Gemini 3 Deep Think, and Gemini Agent, all of which were announced and released last month.

Gemini 3 Flash, now available on Gemini Enterprise, Google Antigravity, Gemini CLI, AI Studio, and on preview in Vertex AI, processes information in near real-time and helps build quick, responsive agentic applications. 

The company said in a blog post that Gemini 3 Flash “builds on the model series that developers and enterprises already love, optimized for high-frequency workflows that demand speed, without sacrificing quality.

The model is also the default for AI Mode on Google Search and the Gemini application. 

Tulsee Doshi, senior director, product management on the Gemini team, said in a separate blog post that the model “demonstrates that speed and scale don’t have to come at the cost of intelligence.”

“Gemini 3 Flash is made for iterative development, offering Gemini 3’s Pro-grade coding performance with low latency — it’s able to reason and solve tasks quickly in high-frequency workflows,” Doshi said. “It strikes an ideal balance for agentic coding, production-ready systems and responsive interactive applications.”

Early adoption by specialized firms proves the model's reliability in high-stakes fields. Harvey, an AI platform for law firms, reported a 7% jump in reasoning on their internal 'BigLaw Bench,' while Resemble AI discovered that Gemini 3 Flash could process complex forensic data for deepfake detection 4x faster than Gemini 2.5 Pro. These aren't just speed gains; they are enabling 'near real-time' workflows that were previously impossible.

More efficient at a lower cost

Enterprise AI builders have become more aware of the cost of running AI models, especially as they try to convince stakeholders to put more budget into agentic workflows that run on expensive models. Organizations have turned to smaller or distilled models, focusing on open models or other research and prompting techniques to help manage bloated AI costs.

For enterprises, the biggest value proposition for Gemini 3 Flash is that it offers the same level of advanced multimodal capabilities, such as complex video analysis and data extraction, as its larger Gemini counterparts, but is far faster and cheaper. 

While Google’s internal materials highlight a 3x speed increase over the 2.5 Pro series, data from independent benchmarking firm Artificial Analysis adds a layer of crucial nuance.

In the latter organization's pre-release testing, Gemini 3 Flash Preview recorded a raw throughput of 218 output tokens per second. This makes it 22% slower than the previous 'non-reasoning' Gemini 2.5 Flash, but it is still significantly faster than rivals including OpenAI's GPT-5.1 high (125 t/s) and DeepSeek V3.2 reasoning (30 t/s).

Most notably, Artificial Analysis crowned Gemini 3 Flash as the new leader in their AA-Omniscience knowledge benchmark, where it achieved the highest knowledge accuracy of any model tested to date. However, this intelligence comes with a 'reasoning tax': the model more than doubles its token usage compared to the 2.5 Flash series when tackling complex indexes.

This high token density is offset by Google's aggressive pricing: when accessing through the Gemini API, Gemini 3 Flash costs $0.50 per 1 million input tokens, compared to $1.25/1M input tokens for Gemini 2.5 Pro, and $3/1M output tokens, compared to $ 10/1 M output tokens for Gemini 2.5 Pro. This allows Gemini 3 Flash to claim the title of the most cost-efficient model for its intelligence tier, despite being one of the most 'talkative' models in terms of raw token volume. Here's how it stacks up to rival LLM offerings:

<…

Konten dipersingkat otomatis.

🔗 Sumber: venturebeat.com

Author: timuna