MAROKO133 Update ai: MIT Researchers Unveil “SEAL”: A New Step Towards Self-Improving AI E

📌 MAROKO133 Breaking ai: MIT Researchers Unveil “SEAL”: A New Step Towards Self-Im

The concept of AI self-improvement has been a hot topic in recent research circles, with a flurry of papers emerging and prominent figures like OpenAI CEO Sam Altman weighing in on the future of self-evolving intelligent systems. Now, a new paper from MIT, titled “Self-Adapting Language Models,” introduces SEAL (Self-Adapting LLMs), a novel framework that allows large language models (LLMs) to update their own weights. This development is seen as another significant step towards the realization of truly self-evolving AI.

The research paper, published yesterday, has already ignited considerable discussion, including on Hacker News. SEAL proposes a method where an LLM can generate its own training data through “self-editing” and subsequently update its weights based on new inputs. Crucially, this self-editing process is learned via reinforcement learning, with the reward mechanism tied to the updated model’s downstream performance.

The timing of this paper is particularly notable given the recent surge in interest surrounding AI self-evolution. Earlier this month, several other research efforts garnered attention, including Sakana AI and the University of British Columbia’s “Darwin-Gödel Machine (DGM),” CMU’s “Self-Rewarding Training (SRT),” Shanghai Jiao Tong University’s “MM-UPT” framework for continuous self-improvement in multimodal large models, and the “UI-Genie” self-improvement framework from The Chinese University of Hong Kong in collaboration with vivo.

Adding to the buzz, OpenAI CEO Sam Altman recently shared his vision of a future with self-improving AI and robots in his blog post, “The Gentle Singularity.” He posited that while the initial millions of humanoid robots would need traditional manufacturing, they would then be able to “operate the entire supply chain to build more robots, which can in turn build more chip fabrication facilities, data centers, and so on.” This was quickly followed by a tweet from @VraserX, claiming an OpenAI insider revealed the company was already running recursively self-improving AI internally, a claim that sparked widespread debate about its veracity.

Regardless of the specifics of internal OpenAI developments, the MIT paper on SEAL provides concrete evidence of AI’s progression towards self-evolution.

Understanding SEAL: Self-Adapting Language Models

The core idea behind SEAL is to enable language models to improve themselves when encountering new data by generating their own synthetic data and optimizing their parameters through self-editing. The model’s training objective is to directly generate these self-edits (SEs) using data provided within the model’s context.

The generation of these self-edits is learned through reinforcement learning. The model is rewarded when the generated self-edits, once applied, lead to improved performance on the target task. Therefore, SEAL can be conceptualized as an algorithm with two nested loops: an outer reinforcement learning (RL) loop that optimizes the generation of self-edits, and an inner update loop that uses the generated self-edits to update the model via gradient descent.

This method can be viewed as an instance of meta-learning, where the focus is on how to generate effective self-edits in a meta-learning fashion.

A General Framework

SEAL operates on a single task instance (C,τ), where C is context information relevant to the task, and τ defines the downstream evaluation for assessing the model’s adaptation. For example, in a knowledge integration task, C might be a passage to be integrated into the model’s internal knowledge, and τ a set of questions about that passage.

Given C, the model generates a self-edit SE, which then updates its parameters through supervised fine-tuning: θ′←SFT(θ,SE). Reinforcement learning is used to optimize this self-edit generation: the model performs an action (generates SE), receives a reward r based on LMθ′’s performance on τ, and updates its policy to maximize the expected reward.

The researchers found that traditional online policy methods like GRPO and PPO led to unstable training. They ultimately opted for ReST^EM, a simpler, filtering-based behavioral cloning approach from a DeepMind paper. This method can be viewed as an Expectation-Maximization (EM) process, where the E-step samples candidate outputs from the current model policy, and the M-step reinforces only those samples that yield a positive reward through supervised fine-tuning.

The paper also notes that while the current implementation uses a single model to generate and learn from self-edits, these roles could be separated in a “teacher-student” setup.

Instantiating SEAL in Specific Domains

The MIT team instantiated SEAL in two specific domains: knowledge integration and few-shot learning.

• Knowledge Integration: The goal here is to effectively integrate information from articles into the model’s weights.
• Few-Shot Learning: This involves the model adapting to new tasks with very few examples.

Experimental Results

The experimental results for both few-shot learning and knowledge integration demonstrate the effectiveness of the SEAL framework.

In few-shot learning, using a Llama-3.2-1B-Instruct model, SEAL significantly improved adaptation success rates, achieving 72.5% compared to 20% for models using basic self-edits without RL training, and 0% without adaptation. While still below “Oracle TTT” (an idealized baseline), this indicates substantial progress.

For knowledge integration, using a larger Qwen2.5-7B model to integrate new facts from SQuAD articles, SEAL consistently outperformed baseline methods. Training with synthetically generated data from the base Qwen-2.5-7B model already showed notable improvements, and subsequent reinforcement learning further boosted performance. The accuracy also showed rapid improvement over external RL iterations, often surpassing setups using GPT-4.1 generated data within just two iterations.

Qualitative examples from the paper illustrate how reinforcement learning leads to the generation of more detailed self-edits, resulting in improved performance.

While promising, the researchers also acknowledge some limitations of the SEAL framework, including aspects related to catastrophic forgetting, computational overhead, and context-dependent evaluation. These are discussed in detail in the original paper.

Original Paper: https://arxiv.org/pdf/2506.10943

Project Site: https://jyopari.github.io/posts/seal

Github Repo: https://github.com/Continual-Intelligence/SEAL

The post MIT Researchers Unveil “SEAL”: A New Step Towards Self-Improving AI first appeared on Synced.

🔗 Sumber: syncedreview.com

📌 MAROKO133 Eksklusif ai: Consultants Forced to Pay Money Back After Getting Caugh

Financial consulting firm Deloitte was forced to reissue the Australian government $291,000 US after getting caught using AI and including hallucinated numbers in a recent report.

As The Guardian reports, Australia’s Department of Employment and Workplace Relations (DEWR) confirmed that the firm agreed to repay the final installment as part of its contract. It had been commissioned in December to review a system that automates penalties in the welfare system in case jobseekers don’t meet their mutual obligations.

However, the “independent assurance review” bore concerning signs that Deloitte had cut corners, and included multiple errors such as references to nonexistent citations — a hallmark of AI slop.

The “hallucinations” once again highlight how generative AI use in the workplace can allow glaring mistakes to slip through, from lawyers getting caught citing nonexistent cases to Trump’s Centers for Disease Control referencing a study that was dreamed up by AI earlier this year.

Deloitte, among other consulting firms, have poured billions of dollars into developing AI tools that they say could speed up their audits, as the Financial Times reports.

Earlier today, the newspaper noted that the United Kingdom’s six largest accounting firms hadn’t been formally monitoring how AI impacts the quality of their audits, highlighting the possibility that many other reports may include similar hallucinations.

University of Sydney sociological lecturer Christopher Rudge, who first highlighted the issues with Deloitte’s DEWR report, said that the company tried to cover its tracks after sharing an updated version of the error-laden report.

“Instead of just substituting one hallucinated fake reference for a new ‘real’ reference, they’ve substituted the fake hallucinated references and in the new version, there’s like five, six or seven or eight in their place,” he told The Guardian. “So what that suggests is that the original claim made in the body of the report wasn’t based on any one particular evidentiary source.”

Despite being caught red-handed using AI to generate hallucinated citations, Deloitte said that the overall thrust of its guidance hadn’t changed. A footnote in the revised version noted that staffers had used OpenAI’s GPT-4o for the report.

“Deloitte conducted the independent assurance review and has confirmed some footnotes and references were incorrect,” a spokesperson told The Guardian. “The substance of the independent review is retained, and there are no changes to the recommendations.”

But outraged lawmakers calling for more oversight.

“Deloitte has a human intelligence problem,” Labor senator Deborah O’Neill, who represents New South Wales, told the Australian Financial Review. “This would be laughable if it wasn’t so lamentable… too often, as our parliamentary inquiries have shown, these consulting firms win contracts by promising their expertise, and then when the deal is signed, they give you whatever [staff] costs them the least.”

“Anyone looking to contract these firms should be asking exactly who is doing the work they are paying for, and having that expertise and no AI use verified,” O’Neill added. “Otherwise, perhaps instead of a big consulting firm procurers would be better off signing up for a ChatGPT subscription.”

“This report was meant to help expose the failures in our welfare system and ensure fair treatment for income support recipients, but instead Labor [is] letting Deloitte take them for a ride,” Greens senator Penny Allman-Payne told the AFR. “Labor should be insisting on a full refund from Deloitte, and they need to stop outsourcing their decisions to their consultant mates.”

More on hallucinations: Fixing Hallucinations Would Destroy ChatGPT, Expert Finds

The post Consultants Forced to Pay Money Back After Getting Caught Using AI for Expensive “Report” appeared first on Futurism.

🔗 Sumber: futurism.com