Search results for "WEB"
2026-04-24
04:54

Perplexity Discloses Web Search Agent Post-Training Method; Qwen3.5-Based Model Outperforms GPT-5.4 on Accuracy and Cost

Perplexity uses SFT followed by RL with Qwen3.5 models, leveraging a multi-hop QA dataset and rubric checks to boost search accuracy and efficiency, achieving best-in-class FRAMES performance. Abstract: Perplexity's post-training workflow for web-search agents combines supervised fine-tuning (SFT) to enforce instruction-following and language consistency with online reinforcement learning (RL) via the GRPO algorithm. The RL stage uses a proprietary multi-hop verifiable QA dataset and rubric-based conversational data to prevent SFT drift, with reward gating and within-group efficiency penalties. Evaluation shows Qwen3.5-397B-SFT-RL achieving top FRAMES performance, 57.3% accuracy with a single tool call and 73.9% with four calls at $0.02 per query, outperforming GPT-5.4 and Claude Sonnet 4.6 on these metrics. Pricing is API-based and excludes caching.
More
03:18

TRADOOR (Tradoor) down 2.20% in 24 hours

Gate News update: On April 23, according to Gate market data, as of the time of publication, TRADOOR (Tradoor) is trading at $8.17. It is down 2.20% over the past 24 hours, with a high of $11.00 and a low of $7.52. The 24-hour trading volume is $7.8239 million. Its current market cap is approximately $117 million. Tradoor is the fastest and simplest way to trade Options and Perps on the web, mobile, and Telegram. You can get started with less capital, enjoy private, one-click trading, with no hidden fees. This is not investment advice; investing involves risks, including market volatility.
More
TRADOOR-29,44%
09:25

Hugging Face Open-Sources ml-intern, an AI Agent for Autonomous ML Research

Open-sourced ml-intern, Hugging Face's autonomous ML research agent that reads papers, curates data, trains, evaluates, and iterates across science, medicine, and math. Abstract: Hugging Face's ml-intern is an open-source autonomous ML research agent that reads papers, curates datasets, trains on local or cloud GPUs, evaluates results, and iterates improvements. Built on smolagents with CLI and web interfaces, it navigates arXiv/HF Papers, HF Hub, and HF Jobs. Demonstrations span science, medicine, and mathematics, showing end-to-end automation and performance gains.
More