📌 TOPINDIATOURS Eksklusif ai: New Gemini 3.1 Pro crushes previous benchmarks, outp

Google has rolled out Gemini 3.1 Pro, the latest update to its flagship AI model, just months after releasing Gemini 3 in November.

The new version enters preview today for developers, enterprises, and consumers, with Google promising stronger reasoning, better coding performance, and improved handling of long documents.

The company says Gemini 3.1 Pro powers the “core intelligence” behind recent upgrades to its Deep Think tool.

While benchmark gains appear modest in some areas, Google claims the update delivers more consistent and reliable performance in real-world tasks.

Stronger reasoning benchmarks

Google highlighted performance gains across several industry tests.

In Humanity’s Last Exam, which measures advanced domain knowledge, Gemini 3.1 Pro scored 44.4 percent. Gemini 3 Pro reached 37.5 percent. OpenAI’s GPT 5.2 scored 34.5 percent.

The company also pointed to a sharp improvement in ARC-AGI-2, a benchmark designed to test novel reasoning problems. Gemini 3 scored 31.1 percent in earlier testing.

Gemini 3.1 Pro jumped to 77.1 percent, more than doubling the previous result.

However, Gemini 3.1 Pro does not top every leaderboard. On Arena, formerly LM Arena, Claude Opus 4.6 leads Gemini in text tasks. It edges Gemini by four points at 1504.

In coding categories, Opus 4.6, Opus 4.5, and GPT 5.2 High also rank ahead.

Arena rankings rely on user voting. Participants choose outputs they prefer.

That format can reward answers that appear correct, even if they contain subtle flaws.

Google designed Gemini 3.1 Pro with developers in mind. The model generates code, explains complex functions, and helps debug errors. It now handles larger code blocks in a single session.

That reduces interruptions during development workflows.

The update also expands long-context capabilities. Gemini 3.1 Pro supports up to one million input tokens and 64,000 output tokens.

Businesses can upload lengthy contracts, reports, or research documents and ask detailed questions without splitting files.

Google kept API pricing unchanged at $2 per million input tokens and $12 per million output tokens.

That stability may appeal to startups and enterprise teams building AI-driven products.

The model also showed gains in the APEX-Agents benchmark, nearly doubling its earlier score.

That benchmark measures performance in agentic workflows, where AI systems execute multi-step tasks.

Enterprise AI push

Google is deploying Gemini 3.1 Pro across its ecosystem. Developers can access it in AI Studio and the Antigravity IDE.

Enterprise customers will see it in Vertex AI and Gemini Enterprise. Consumers can use it through the Gemini app and NotebookLM.

The company says it improved safety controls and monitoring systems.

Businesses handling sensitive data demand stable and predictable outputs.

Google aims to position Gemini 3.1 Pro as a dependable tool for customer support, automation, and document review.

The broader AI market continues to accelerate in the United States. Companies now compare models on reasoning strength, coding depth, and long-context performance.

Gemini 3.1 Pro may not dominate every leaderboard, but Google appears focused on practical gains that matter inside real workflows.

If past patterns continue, Google could soon release a 3.1 update for its faster and lower-cost Flash model.

For now, Gemini 3.1 Pro signals Google’s intent to compete aggressively in enterprise AI.

🔗 Sumber: interestingengineering.com

📌 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