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

📌 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 Hot ai: US lab doubles high-voltage sodium battery lifespans, cell pro

Researchers at the Pacific Northwest National Laboratory have developed a meta-weakly solvating electrolyte that allows for the stable operation of high-voltage sodium-ion batteries.

“Developing alternative battery systems based on earth-abundant elements has thus become increasingly important,” said the researchers in a new study published in the journal Nano Energy.

“Given that sodium is the closest alkali metal to lithium, sharing similar chemistry while being far more available, sodium-ion batteries (SIBs) are naturally positioned as a promising technology for next-generation energy storage.”

In laboratory tests, full cells using this electrolyte retained 80 percent of their initial capacity after 500 cycles. This performance level is higher than that of standard benchmark devices, which typically sustain between 100 and 300 cycles before reaching similar levels of degradation.

“The full cells demonstrate 80% capacity retention after 500 cycles, outperforming both conventional carbonate-based and localized high-concentration electrolytes,” added the study.

The electrochemical testing was conducted at a constant temperature of 30 degrees Celsius using sodium hexafluorophosphate and sodium bis(fluorosulfonyl)imide salts.

Post-cycling analysis to evaluate condition

The researchers also performed post-cycling analysis after 50 cycles, using scanning electron microscopy and energy-dispersive X-ray spectroscopy to evaluate the condition of the electrodes.

The results demonstrated that the new electrolyte formulation improved high-voltage interfacial stability and reduced leakage current when paired with sodium nickel manganese iron oxide cathodes and hard carbon anodes.

Most conventional battery electrolytes are designed to strongly solvate metal ions to assist their movement through the liquid. This process creates a stable ion-solvent shell that can be difficult to break apart when the ion reaches the electrode surface.

When the shell does not detach properly, electrolyte molecules are often pulled into unwanted side reactions at the interface. These reactions form unstable layers and consume the electrolyte, which leads to the gradual degradation of the battery cell materials over time.

The design from the Pacific Northwest National Laboratory utilizes an intermediate solvation structure where sodium ions are less tightly bound to solvent molecules.

“We discovered that replacing conventional non-solvating diluents in LHCEs, such as 1,1,2,2-tetrafluoroethyl 2,2,3,3-tetrafluoropropylether (TTE), with the weakly solvating tris(2,2,2-trifluoroethyl) phosphate (TFP) can address the abovementioned limitations of LHCEs while maintaining an anion-rich environment around sodium ions,” explained the researchers.

Casting a slurry onto aluminium foil

The electrodes were constructed by casting a slurry onto aluminum foil using binders such as polyvinylidene fluoride, sodium carboxymethyl cellulose, and styrene-butadiene rubber, along with conductive carbon additives.

Scientists evaluated the battery’s performance using nuclear magnetic resonance spectroscopy to analyze the specific solvation structures and their behavior at the electrode interface.

These tests showed that the meta-weakly solvating electrolyte allowed for faster sodium desolvation and lower charge-transfer resistance compared to conventional options. Lead author An L. Phan stated that this strategy regulates the sodium solvation structure to facilitate favorable reactions while suppressing unwanted ones.

The result is a reduction in irreversible material loss and improved electrochemical stability during extended cycling.

“The new electrolyte represents a new strategy to regulate Na solvation structure that can facilitate favorable reactions and suppress unwanted ones,” the research’s lead author An L. Phan told ESS News.

“This results in reduced irreversible loss and degradation of cell materials under practical conditions.”

🔗 Sumber: interestingengineering.com