accuracy

Researchers at the Chinese People's Public Security University have developed an AI framework capable of detecting illegal cryptocurrency transactions with an overall accuracy rate close to 90%. The study was published in the Chinese peer-reviewed journal "Journal of Intelligence," and the corresponding author, Dr. Sun Jingchao (specializing in criminal investigation and cybersecurity), noted that the research "provides an accurate, scalable, and interpretable solution for detecting illegal cryptocurrency transactions," and offers "an innovative technical path" for regulatory agencies to combat illegal cryptocurrency transactions and economic crimes.This AI framework utilizes memory modules and large language models, specifically targeting the anonymity and cross-border characteristics of cryptocurrencies to track illegal activities such as money laundering. The release of this research coincides with China's ongoing efforts to intensify the crackdown on financial crimes related to cryptocurrencies. In March of this year, the Supreme People's Procuratorate of China disclosed that by 2025, procuratorial authorities had prosecuted 3,259 individuals for money laundering crimes involving virtual currencies and underground banks.As the trading volume of cryptocurrencies rapidly increases, their anonymity and cross-border characteristics provide a channel for illegal fund flows. This police-developed AI detection tool marks a shift in regulatory technology from passive tracking to proactive intelligent identification.

Fireworks AI released an evaluation report comparing the open-source model Kimi K3 and the closed-source model Fable 5 across approximately 1030 agent tasks (covering scenarios such as SWE, terminal operations, algorithms, multilingual coding, and legal matters). The results showed that the accuracy rates for the two models on the SWE benchmark were 92.4% and 92.6%, respectively, indicating similar overall performance, but each model has its strengths in specific areas: K3 leads in symbolic mathematics and development tools, while Fable excels in web and data visualization; in long-cycle terminal tasks, K3 independently tackled 11 tasks that Fable could not solve.The report pointed out that adopting a task-level routing strategy can dynamically allocate between the two, achieving an accuracy rate of 93%, surpassing either single model. Additionally, due to K3's significant cost advantage on the Fireworks platform, costs for long-cycle agent tasks can be up to 50 times lower than Fable. The report recommends using the open-source model as the default option and continuously learning the optimal match between tasks and models through routers to balance quality and cost.

According to official data from NeoSoul, during the 2026 World Cup, the platform had approximately 28,000 AI Agents participating in predictions, completing a total of 294,000 predictions, with an overall prediction accuracy rate of 82%. Among them, 1,111 Agents maintained a perfect record after completing at least 10 predictions.It is reported that NeoSoul is the largest AI Agent cluster infrastructure on the BNB Chain and 0G ecosystem, dedicated to supporting Agents' autonomous decision-making, participating in real transactions, and promoting the development of the AI Economy network.

The largest AI Agent cluster infrastructure of BNB Chain and the 0G ecosystem, EvoEvo under NeoSoul, has accurately predicted 8 matches in this World Cup, achieving an overall accuracy rate of 70%.It is reported that EvoEvo is committed to forming collective intelligence by aggregating AI Agent judgments, supporting Agent autonomous decision-making and trading, and promoting the development of the AI economy.

According to CoinDesk, as the U.S. Congress failed to complete the full legislative process for the funding bill, the U.S. government is expected to enter a short-term, partial shutdown starting Saturday Eastern Time. However, this event also highlights the issue of "definition precision" in prediction markets. The prediction contracts on Polymarket and Kalshi regarding "whether the U.S. government will shut down" show differences due to varying definitions of triggering conditions. The relevant contracts on Polymarket indicate a shutdown probability of about 88%, while similar contracts on Kalshi have a probability as high as approximately 93%. This discrepancy may be because some contracts consider the formal announcement of a shutdown by the Office of Personnel Management (OPM) as the basis for settlement, meaning that even a partial shutdown is regarded as valid. This event underscores the sensitivity of prediction markets to contract triggering conditions, official confirmation entities, and timelines in macro events, with the details of the contracts themselves becoming an important factor influencing the odds.

ChainCatcher news, according to Jinshi reports, U.S. Treasury Secretary Basant stated that if the data is accurate, the Federal Reserve may consider cutting interest rates earlier.

ChainCatcher news, Messari released a research report on the decentralized AI infrastructure Mira. Mira optimizes the reliability of AI outputs through a distributed model consensus mechanism, with its verification layer improving the accuracy of AI facts in scenarios such as finance and education from 70% to 96%. The protocol breaks down AI outputs into independent factual statements, which are cross-verified by heterogeneous models provided by node operators such as Io.Net and Aethir, requiring consensus from over 2/3 of the nodes to pass.Mira currently processes over 3 billion text tokens daily, covering 4.5 million users across chatbots, educational platforms, and more. The protocol employs an economic incentive model, where verification nodes receive rewards based on their contributions, while anomalous nodes face penalties. Partners include decentralized GPU computing providers like Hyperbolic and Exabits, achieving computing power expansion through a node delegation mechanism.According to team data, the protocol has reduced the AI hallucination rate by 90%, with each verification taking less than 30 seconds. Users can trace the verification process through on-chain proof, with each output accompanied by an encrypted certificate recording model voting details. Currently, integrated applications like Klok have utilized this technology to optimize educational content generation, with plans to expand into high-risk areas such as medical diagnosis in the future.

ChainCatcher news, data scientist Alex McCullough's research shows that the prediction market Polymarket has an accuracy rate of 90% when predicting events occurring a month later, and an accuracy rate of up to 94% four hours before the event occurs. McCullough analyzed Polymarket's historical data and found that after removing extreme probability values, the platform slightly but consistently overestimates the probability of events occurring within most probability ranges, possibly influenced by factors such as herd mentality, low liquidity, and participants' preference for high-risk bets.

ChainCatcher message, Bitget's flash exchange feature has been upgraded. This optimization further enhances the accuracy of quotes and significantly increases the single transaction limit.After the upgrade, the trading range for regular currency pairs in the flash exchange has been adjusted to 2 USDT - 50,000 USDT, and the maximum exchangeable amount for mainstream currency pairs like BTC and ETH is now 300,000 USDT, providing users with a more flexible and efficient trading experience. Please update to the latest version to experience the optimized features.

ChainCatcher news, according to CoinDesk, Swedish blockchain aviation data network company Wingbits will launch satellites via the SpaceX Transporter-13 mission at the end of February to enhance the data verification capabilities of its decentralized physical infrastructure network (DePIN).The company's CEO Robin Wingardh stated at the Hong Kong Consensus conference that over 2,200 ground nodes have been deployed in more than 90 countries worldwide, tracking approximately 120,000 unique flights daily.Wingbits uses Solana testnet tokens to incentivize data providers, differentiating itself from traditional aviation data companies FlightAware and Flightradar24. The company recently completed a $5.6 million second seed funding round backed by investors including Borderless Capital and Bullish Capital.

ChainCatcher news, according to CoinDesk, regarding the previous claim by the metaverse platform Decentraland that the blockchain data provider DappRadar's statistics are inaccurate, DappRadar stated on its official blog that it is recalculating Decentraland's user data. Initially, DappRadar only tracked 13 Decentraland contracts, but they will now track over 3,000 contracts and will begin exploring methods to include metaverse transactions in their tracking.According to DappRadar, the statistics will track the number of unique active wallets (UAW), the number of transactions, and the transaction volume across dapps. Typically, they do not track any dapp activity outside the blockchain ecosystem, which may lead to discrepancies between their measurements and those reported by the dapps themselves.ChainCatcher previously reported that DappRadar data showed Decentraland had only 38 active users in the past 24 hours. Decentraland responded that some websites only track specific smart contract transactions but inaccurately report them as daily active users (DAU). According to their statistics, the monthly active users in September reached 56,697. (CoinDesk)