📌 TOPINDIATOURS Update ai: MIT Researchers Unveil “SEAL”: A New Step Towards Self-

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

📌 TOPINDIATOURS Hot ai: Removing old sofas can halve cancerous flame retardants in

A decade-long study has shown that household environments play a significant role in long-term chemical exposure, and that removing older foam-based furniture from homes can halve levels of toxic flame retardants in the human body.

The study, carried out by the California Department of Public Health along with environmental health organizations such as the Environmental Working Group (EWG), tracked blood and household dust samples from dozens of participants over 10 years.

According to EWG, the project examined two major classes of flame retardants, polybrominated diphenyl ethers (PBDEs) and organophosphate flame retardants (OPFRs) that have been used in furniture like chairs and sofas.

PBDE exposure has previously been tied to cancer, neurotoxicity, thyroid disease, and reproductive harm. The new study found that PBDE levels dropped up to four times faster in people who removed older furniture from their homes.

Toxins in furniture

PBDEs were widely used in upholstered furniture, electronics, and consumer products from the 1970s through the mid-2010s to meet flammability standards, particularly in California.

These toxic chemicals were added to polyurethane foam and were not chemically bound. This allowed them to migrate over time and build up in household dust that people can inhale or ingest.

PBDEs can also be absorbed through the skin, with infants and young children at greatest risk. “Infants and young children are especially at risk since they crawl and play on the floor, where contaminated dust settles, and then frequently put their hands in their mouths,” EWG said.

To determine whether removing older furniture could reduce exposure to PBDEs, scientists from the California Department of Public Health joined forces with EWG, the California Environmental Protection Agency, Silent Spring, the Green Science Policy Institute, the Sequoia Foundation, and the University of California, Davis.

The collaboration, which launched in 2015, focused on a few dozen households. It examined how removing furniture manufactured before 2014 affected participants’ levels of flame retardants.

“In 2013, the Golden State updated its furniture flammability standards, no longer requiring the use of chemical flame retardants,” EWG noted. “Study participants collected household dust before and after removing a couch or couch foam.”

Home sources matter

The most recent blood analyses found no significant change in OPFR levels after furniture removal, suggesting that exposure also comes from other sources, such as electronics and vehicles.

However, PBDE levels fell two to four times faster in participants who removed older furniture than in those who did not, with blood concentrations dropping by about half within roughly 1.4 years.

“Furniture made today is less likely to contain harmful flame retardants,” EWG concluded in a statement. “But older furniture continues to be a possible source of exposure, even years after it was manufactured.”

The researchers recommended replacing old furniture with products that do not contain these chemicals. They emphasized that changing out couch cushion foam is a cheaper alternative, and that regular dusting and HEPA vacuuming can help lower flame-retardant levels when full furniture removal isn’t feasible.

The study has been published in the journal Environmental Pollution.

🔗 Sumber: interestingengineering.com