📌 TOPINDIATOURS Breaking ai: MIT Researchers Unveil “SEAL”: A New Step Towards Sel

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 Eksklusif ai: Doctor Says What Border Patrol Agents Did After Shoo

A doctor who witnessed the brutal shooting of Alex Pretti by Border Patrol agents says that the agents didn’t even bother trying to save the victim’s life as he bled out on the ground — because they were too busy counting the bullet wounds.

The harrowing testimony, detailed in a sworn court filing, contradicts key details in the accounts given by the Department of Homeland Security and top officials in the Trump administration, which insist that the 37-year-old ICU nurse threatened agents and attempted to reach for his gun during an altercation in Minneapolis on Saturday.

According to the witness, a 29-year-old physician who isn’t identified by name in the court document, Pretti was yelling at the agents but did not attack them or brandish any sort of weapon before the situation escalated fatally.

“Suddenly, an ICE agent shoved him to the ground,” the witness said, referring to the Border Patrol officers. “My view of the altercation was partially obstructed, but after a few seconds, I saw at least four ICE agents point guns at the man. I then saw the agents shoot the man at least six or seven times.”

When he rushed to the scene and told Border Patrol he was a physician and wanted to attend to the victim, the agents demanded to see his physician’s license — “which I obviously didn’t have,” he said.

Shockingly, he noticed that “none of the ICE agents who were near the victim were performing CPR” despite him clearly being in “critical condition,” so the doctor continued his protests.

“Normally, I would not have been so persistent, but as a physician, I felt a professional and moral obligation to help this man,” he said, “especially since none of the agents were helping him.”

Once the agents finally let him through, the first red flag that jumped out at him was that the victim was lying on his side.

“That is not standard practice when a victim has been shot. Checking for a pulse and administering CPR is standard practice,” he said. “Instead of doing either of those things, the ICE agents appeared to be counting his bullet wounds.”

The doctor checked for a pulse, found that Pretti had none, and immediately began CPR. According to the physician, Pretti had at least three bullet wounds in his back, another on his upper left chest, and possibly one on his neck. The doctor left after EMS arrived at the scene and took charge.

He described how the scene had left him shaken afterwards.

“When I returned to my apartment, I was extremely distraught,” the doctor said. “I was sobbing and shaking uncontrollably.”

Soon, tear gas began to flood into his apartment from a protest outside, and he left to drive to a friend’s house, “still crying and shaking” and “barely able to speak.”

Pretti’s killing has added to the intense backlash against ICE and Border Patrol’s brutal crackdown in Minneapolis, where the federal agents have clashed with onlookers, rounded up both citizen and non-citizen children, and shot and killed another protestor, 37-year-old Renee Good.

Contradicting the mountains of video evidence taken at the scene, the DHS maintains that Pretti, a nurse that treated sick and wounded veterans at the Minneapolis VA Medical Center, “wanted to do maximum damage and massacre law enforcement.” Agency secretary Kristi Noem called his actions “domestic terrorism,” and White House Deputy chief of staff Stephen Miller accused him of being an “assassin.”

The doctor wholeheartedly disagrees.

“The victim was not actively threatening ICE agents or the public — he was just yelling at the agents because he objected to ICE’s presence in our city,” he said in the filing.

“I am not sure when I will return to my apartment,” he added. “I do not feel safe in my city.”

More on immigration: ICE Is Scanning Civilians’ Faces, Telling Them They’re Being Entered Into a Terrorism Database

The post Doctor Says What Border Patrol Agents Did After Shooting Alex Pretti Was Sickening appeared first on Futurism.

🔗 Sumber: futurism.com