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

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

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: YouTuber builds 7 times larger Arduino tortoise bot that

Instead of sticking with the typical tabletop-scale builds, Arduino hobbyist and YouTuber UncleStem chose to take a familiar robotics concept and push it into much larger territory. 

After completing a custom Arduino Uno board that was scaled up seven times its usual size, he set out to find a project that could properly test the oversized controller in a real-world application. He landed on a turtle-style obstacle-avoiding robot, a design known for its simplicity and reliability.

The original version relies on basic sensor input and straightforward movement logic, making it an ideal candidate for scaling. By increasing every component by a factor of seven, the project transforms a compact educational robot into a significantly larger machine while preserving the same fundamental behavior and control principles.

Custom motor shells help scale up design 

To handle the increased size and weight of the project, the build started with a shift in hardware strategy. Instead of relying on the small hobby motors typically used in basic Arduino turtle bots, UncleStem turned to motors salvaged from children’s ride-on toys. These 24-volt units deliver significantly more torque, making them suitable for driving a much larger chassis, TechEBlog writes.

Rather than simply mounting them as-is, he designed custom-fit shells that slide over the motors, maintaining a clean, unified look. This approach preserves the visual identity of the original compact robot while adapting it to a far more demanding mechanical scale. The result is a system that combines practical power with a carefully considered, scaled-up aesthetic.

Because of the size and precision required for the frame, the laser-cut components couldn’t be handled with standard workshop equipment. UncleStem opted to outsource the cutting process to a professional service equipped to work with large-format sheets. The parts were produced from a full-sized 1mm acrylic panel, ensuring clean edges and consistent accuracy across the oversized structure.

Ordinary workshop cutters would have struggled with both the scale and material constraints, making outsourcing the most reliable option for maintaining structural integrity and dimensional precision in the build.

Lawn equipment wheels and metal rods help power robot

At the base of the build, the wheels sourced from lawn equipment suppliers provide surprisingly smooth movement under the robot’s heavy frame. To maintain visual consistency, 3D-printed hubcaps were added, helping unify the oversized design.

Control is handled by a custom-built, enlarged Arduino Uno created by UncleStem, which conceals a standard Arduino Nano inside for actual computation and code execution. Even the wiring had to be improvised at scale, with metal rods used in place of traditional jumper wires, since no off-the-shelf equivalents exist for a project of this size.

The navigation logic follows the classic turtle-bot pattern without modification. The robot drives forward in a straight line until its front sensors detect an obstacle, triggering an immediate stop. It then performs a sweep by checking left, center, and right to evaluate available space, identifies the clearest path, and executes a turn in that direction before resuming forward motion. 

This simple decision loop repeats continuously, allowing the system to operate autonomously using only basic obstacle detection and directional comparison.

🔗 Sumber: interestingengineering.com