MAROKO133 Breaking ai: AI “Research” Papers Are Complete Slop, Experts Say Hari Ini

📌 MAROKO133 Update ai: AI “Research” Papers Are Complete Slop, Experts Say Edisi J

There’s sloppy science, and there’s AI slop science.

In an ironic twist of fate, beleaguered AI researchers are warning that the field is being choked by a deluge of shoddy academic papers written with large language models, making it harder than ever for high quality work to be discovered and stand out.

Part of the problem is that AI research has surged in popularity. The more people who jump on the wagon, the more some are trying to speedrun an academic reputation by churning out dozens — and sometimes even hundreds — of papers a year, giving the entire pursuit a bad name.

In an interview with The Guardian, professor of computer science at UC Berkeley Hany Farid called the state of affairs a “frenzy.” With so much slop rising to the top, he says he now advises his students not to enter the field.

“So many young people want to get into AI,” Farid told The Guardian. “It’s just a mess. You can’t keep up, you can’t publish, you can’t do good work, you can’t be thoughtful.”

Farid stirred debate over the topic by calling out the output of an AI researcher named Kevin Zhu, who claims to have published 113 papers on AI this year.

“I can’t carefully read 100 technical papers a year,” Farid wrote in a LinkedIn post last month, “so imagine my surprise when I learned about one author who claims to have participated in the research and writing of over 100 technical papers in a year.”

Zhu, who recently received his bachelor’s in computer science at UC Berkeley — the same place that Farid teaches —launched an AI researcher program aimed at high schoolers and college students called Algoverse. Many of its participants are coauthors on Zhu’s papers, The Guardian noted. Each student pays $3,325 for a 12-week online course, during which they’re expected to submit work to AI conferences.

One of those conferences is NeurIPS, which is considered to be one of the big three conferences in a field that was once obscure but is now the center of attention as AI commands immense investment and social cachet. In 2020 it fielded less than 10,000 papers, according to The Guardian. This year, that number has jumped to over 21,500, a trend shared by other major AI conferences. The explosion has been so extreme that NeurIPS is now relying on PhD students to help review its flood of submissions.

The overwhelming volume is thanks to people like Zhu: 89 of his over a century of papers are being presented at NeurIPS this week.

Farid called Zhu’s papers a “disaster,” and added that he “could not have possibly meaningfully contributed” to them.

“I’m fairly convinced that the whole thing, top to bottom, is just vibe coding,” Farid said using the new slang that’s emerged to describe using AI tools to quickly build software, exemplifying the attitude of reckless abandon that the new crop of AI-dependent programmers are taking to the practice.

Zhu would not confirm or deny whether his papers were written with AI when asked by The Guardian, but said his teams used “standard productivity tools such as reference managers, spellcheck, and sometimes language models for copy-editing or improving clarity.”

The role that AI has rapidly carved out in academic research has been a point of controversy ever since it first surged in popularity several years ago. Tools like ChatGPT are still prone to hallucinating citations, or inventing sources that do not exist, which often sneak through the peer review process of even prestigious journals. Other instances, such as when a peer-reviewed paper used an AI-generated diagram of a mouse with impossibly super-sized genitalia, make you question if there’s any oversight at all. The tech is so entrenched in academia that some clever authors are inserting hidden text into their papers designed to trick “reviewers” that are themselves AI-powered into giving positive assessments of their work. 

What’s particularly disconcerting to hear now, however, is how AI research is beginning to be torn apart by the technology itself. How long can the pursuit survive its own product? And what does that mean for the upcoming generation of AI scientists, if novel research is being drowned out by their far more prolific peers that are churning out studies with fabricated sources?

Even a seasoned vet like Farid says it’s now makes it impossible to keep track of what’s happening in the AI field.

“You have no chance, no chance as an average reader to try to understand what is going on in the scientific literature,” Farid told The Guardian. “Your signal-to-noise ratio is basically one. I can barely go to these conferences and figure out what the hell is going on.”

More on AI: AI Researchers Say They’ve Invented Incantations Too Dangerous to Release to the Public

The post AI “Research” Papers Are Complete Slop, Experts Say appeared first on Futurism.

🔗 Sumber: futurism.com

📌 MAROKO133 Eksklusif ai: Researchers from PSU and Duke introduce “Multi-Agent Sys

Share My Research is Synced’s column that welcomes scholars to share their own research breakthroughs with over 2M global AI enthusiasts. Beyond technological advances, Share My Research also calls for interesting stories behind the research and exciting research ideas. 

Meet the author
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

2. failure-responsible agent and the decisive error step that led to the task’s failure.

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 m…

  Konten dipersingkat otomatis.

  🔗 Sumber: syncedreview.com

  ---