Baidu Openly Releases Wenxin Unlimited OCR 2 Document Parsing Model, Supporting "Infinite" Number of Pages
The Block
1h ago
Ai Focus
Baidu officially released and open-sourced the Wenxin Unlimited OCR 2 model on the 27th of this month, claiming that it has increased the document parsing limit from dozens of pages to "infinite," allowing for the simultaneous parsing of over 3,600 pages from the entire series of 7 books in the "Harry Potter" franchise. Baidu stated that this model has achieved end-to-end performance SOTA in both multi-page document parsing and the parsing of ultra-long documents exceeding a hundred pages.
Helpful
No.Help

News from the IT community on October 30th: Baidu ( Baidu ) officially released and open-sourced the Wenxin Unlimited OCR 2 model on the 27th of this month. This model has increased the maximum document parsing capacity from dozens of pages to "infinite," allowing for the simultaneous parsing of over 3,600 pages from the entire series of 7 books in the "Harry Potter" franchise.

Unlimited OCR Natively supports the segmentation and recognition of layout elements, and has the capability to merge cross-page copywriting and tables. It can process more content faster and more efficiently with the same computing power.

Baidu stated that Wenxin Unlimited OCR 2 achieved a score of 84.6 in the MPDocBench benchmark test for multi-page document parsing, and a score of 77.1 in the MarathonDocBench benchmark test specifically evaluating the ability to parse ultra-long documents of over a hundred pages, both achieving end-to-end performance SOTA.

refer to:

https :// ai.baidu.com / support / news ? action = detail & id =3290

Tip
$0
Like
0
Save
0
Views 24
HQYC reminds readers to view blockchain rationally, stay aware of risks, and beware of virtual token issuance and speculation. All content on this site represents market information or related viewpoints only and does not constitute any form of investment advice. If you find sensitive content, please click“Report”,and we will handle it promptly。
Submit
Comment 0
Hot
Latest
No comments yet. Be the first!
Related
Iran claims four oil tankers have hit mines in the Strait of Hormuz; Houthi militants claim responsibility for attacking Saudi Arabia
Iran claims four oil tankers have hit mines and caught fire in the Strait of Hormuz, while Yemen's Houthi rebels claim to have launched missile attacks on Saudi Arabia, further intensifying regional tensions. Meanwhile, reports from outside suggest that the United States is considering resuming military actions against Tehran after the mid-term elections, while Vice President JD Vance is attempting to promote dialogue with Iran.
CNBC
·2026-10-30 15:14:00
9
NVIDIA to Release Financial Reports for the Third Quarter of Fiscal Year 2027 on November 17, Pacific Time
NVIDIA announces that it will hold a conference call at 2 p.m. Pacific Time on November 17 to announce and interpret the financial results for the third quarter of fiscal year 2027. The company will also release the financial results and written comments from Chief Financial Officer Colette Kress in advance.
The Block
·2026-10-30 15:13:58
11
Report: Maiwell Technology is set to acquire optical module manufacturer Lumentum; the deal could be announced next week at the earliest.
According to Bloomberg, which cited people familiar with the matter, Mellanox Technologies is in advanced negotiations to acquire Lumentum, a manufacturer of optical components for data centers. The deal could be announced as early as next week, but no final agreement has been reached yet between the two parties. Following the announcement of the news, Mellanox Technologies' shares fell by more than 8% after the market closed, while Lumentum's shares rose by 11%.
Wallstreetcn
·2026-10-30 14:53:14
26
Blockchain.ai Launches Solana Automatic Copy-Trading Service, allowing replication of operations from any wallet
Blockchain.ai indicates that an automatic trailing trading function for Solana has been launched. Users can select the target wallet, set the position size and limits, and automatically follow its buying and selling through an independent treasury. The company states that this function supports viewing the wallet's historical performance, setting stop-loss points and buy limits, and allows users to fund the treasury using USDC, USDT, or SOL.
U.Today
·2026-10-30 14:43:56
26
Novartis is crossing a patent cliff, but another one is already approaching.
Novartis bets on the experimental multiple sclerosis drug remibrutinib, with peak annual sales expected to exceed $4 billion, but Barclays has lowered its forecasts, believing that the drug's efficacy in relapsing-remitting multiple sclerosis is not as good as analysts initially anticipated. At the same time, Novartis also faces patent pressures for Kisqali and Kesimpta which are due to expire around 2032.
CNBC
·2026-10-30 14:33:58
25
View More