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

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

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: US to boost production of submarine-detection devices th

A Massachusetts-based company is set to start production of devices that are used to detect submarines underwater. Ultra Maritime has received a contract from the U.S. Navy to begin producing the AN/SSQ-125B sonobuoys.

These devices are small, expendable sensors dropped into the ocean—usually from aircraft—that act like temporary underwater listening systems, capturing sound signals and transmitting them back for analysis.

Designed to detect quieter, more advanced submarines

Reports have revealed that the newer SSQ-125B version is designed to detect quieter and more advanced submarines at greater distances and in complex ocean environments, making it a significant upgrade for modern anti-submarine warfare operations.

The company has been awarded a Sole Source Firm Fixed Priced contract for the AN/SSQ-125B Low Rate Initial Production (LRIP) for the U.S. Navy in support of annual training, peacetime operations and testing expenditures, as well as to maintain sufficient inventory to support the execution of major combat operations based on naval munitions requirements process.

Anti-submarine warfare (ASW) technology

“This award marks a significant milestone for Ultra Maritime in the advancement of anti-submarine warfare (ASW) technology for the U.S. Navy. As adversary undersea threats grow increasingly stealthy and technologically sophisticated, the Q-125B is designed to enable detection of even the quietest submarines at greater ranges than ever before,” said Carlo Zaffanella, President and CEO of Ultra Maritime.

“This enhanced performance and extended detection area delivers substantial improvements to the ASW kill chain, providing operators with clearer acoustic data and faster, more confident decision-making.”

Emphasis on improving submarine detection capabilities

The contract allows the Navy to begin producing and deploying these sonobuoys for training, operational use, and inventory buildup. The announcement reflects a growing emphasis on improving submarine detection capabilities as underwater threats become more advanced. It also highlights Ultra Maritime’s role as a key supplier of next-generation sonar and anti-submarine technologies.

As the undersea battlespace becomes increasingly complex, the need for greater resilience against challenging acoustic environments is critical. The Q-125B incorporates advanced signal processing and improved acoustic performance to operate effectively in demanding ocean conditions. To bring this new capability forward, Ultra Maritime has made strategic internal investments in cutting-edge sonar technologies and purpose-built manufacturing facilities dedicated to the development and production of next-generation sonobuoys, according to a press release.

These investments ensure production readiness, scalability to meet growing demand, and long-term sustainment of critical ASW capabilities. Backed by deep domain expertise in acoustic engineering and manufacturing excellence, Ultra Maritime continues to lead the evolution of sonobuoy technology, according to the company.

Reports have revealed that traditional detection methods are becoming less effective as submarine technology advances, so systems like the AN/SSQ-125B are designed to improve sensitivity, signal processing, and performance in challenging acoustic conditions such as shallow waters or areas with high background noise (like shipping lanes or rough seas).

These sonobuoys typically work as part of a larger network, where multiple buoys are deployed over a wide area to create an underwater surveillance grid, allowing operators to triangulate and track submarine movements more accurately, as per reports.

🔗 Sumber: interestingengineering.com