BTC $80,072.85 -0.49%
ETH $2,317.39 -1.03%
BNB $741.70 +0.02%
XRP $1.29 -0.46%
SOL $101.63 +0.35%
TRX $0.3304 +0.37%
DOGE $0.0774 +0.58%
ADA $0.2055 +0.46%
BCH $278.48 +1.10%
LINK $11.48 +0.71%
HYPE $78.33 +1.97%
AAVE $135.88 -0.05%
SUI $0.9145 -0.11%
XLM $0.1690 -0.03%
ZEC $1,236.93 +2.25%
AAPL $344.89 -0.60%
AMZN $249.16 +0.02%
GOOGL $319.95 -0.27%
MSFT $531.19 +1.46%
META $643.09 +0.07%
NVDA $215.86 +1.52%
TSLA $394.74 +6.01%
SNDK $1,263.55 +0.94%
INTC $91.17 +4.57%
SPCX $149.13 +0.17%
MU $907.62 +1.77%
AMD $530.60 +1.52%
BTC $80,072.85 -0.49%
ETH $2,317.39 -1.03%
BNB $741.70 +0.02%
XRP $1.29 -0.46%
SOL $101.63 +0.35%
TRX $0.3304 +0.37%
DOGE $0.0774 +0.58%
ADA $0.2055 +0.46%
BCH $278.48 +1.10%
LINK $11.48 +0.71%
HYPE $78.33 +1.97%
AAVE $135.88 -0.05%
SUI $0.9145 -0.11%
XLM $0.1690 -0.03%
ZEC $1,236.93 +2.25%
AAPL $344.89 -0.60%
AMZN $249.16 +0.02%
GOOGL $319.95 -0.27%
MSFT $531.19 +1.46%
META $643.09 +0.07%
NVDA $215.86 +1.52%
TSLA $394.74 +6.01%
SNDK $1,263.55 +0.94%
INTC $91.17 +4.57%
SPCX $149.13 +0.17%
MU $907.62 +1.77%
AMD $530.60 +1.52%

gen

All
Article
Flash

first_img Google DeepMind released XYEval, assessing whether agents comply with misleading suggestions

Google DeepMind researchers Zhengxuan Wu, Yuxuan Li, Oyvind Tafjord, and Been Kim released a preprint paper titled "XYEval: Agents say yes to bad advice," proposing a meta-evaluation framework XYEval to assess the performance of agents when faced with misleading suggestions from users. The paper extends the "XY problem" to the agent scenario, where the user inquires about their attempted solution X rather than the actual problem Y.XYEval retains the original task environment and verifiable answers, only adding seemingly reasonable but misleading user suggestions in the instructions. The misleading suggestions are generated by a generator after seeing the complete task and are appended in the form of confident suggestions rather than hard requirements, allowing the original task to remain solvable. The evaluation covers six benchmarks: τ2-bench, SWE-bench Verified, SWE-bench Pro, Terminal-Bench, Humanity's Last Exam, and MCP-Atlas, along with five models, where the misleading suggestions for τ2-bench are generated based on rules set by human experts.The paper states that the performance significantly declines after the XY mutation, with the relative drop reaching up to 46.7%. On τ2-bench, when faced with users who require a detailed explanation before accepting a better solution, the performance drops even more. Encouraging models to recognize the XY problem through simple system instructions only partially alleviates the issue. The related code is published in the GitHub repository google-deepmind/xyeval.

YZi Labs launched the first phase of the Atlas Scout program to assist the new generation of college student entrepreneurs in obtaining early support

According to official news, YZi Labs announced the launch of the first phase of the Atlas Scout program, recruiting four students from Stanford University, Yale University, the University of California, Berkeley, and the University of California, Los Angeles, to participate in early investment projects in the fields of Web3, artificial intelligence (AI), and biotechnology.The program aims to allow outstanding university students to gain early exposure to venture capital, participate in project discovery, preliminary research, due diligence, and investment judgment, and encourage them to organize entrepreneurial communities on campus and explore investment opportunities. YZi Labs will provide a maximum of $1 million in intended investment for related investment opportunities in this program and open its ecosystem network composed of over 300 portfolio companies.YZi Labs stated that many potential entrepreneurs often begin their explorations in laboratories, open-source projects, campus entrepreneurial organizations, and technology communities before entering traditional financing channels, and their peers who learn and work with them usually discover their potential earlier. The Atlas Scout program hopes to leverage students' own campus networks to help these early entrepreneurs obtain capital, mentorship, and industry resources. Eligible entrepreneurial teams will also have the opportunity to connect with YZi Labs' EASY Residency program or further discuss investment collaboration.The Atlas Scout program not only expands financing and resource channels for early entrepreneurs but also allows participating students to accumulate venture capital experience through practical projects, enhancing their project evaluation, founder communication, and investment judgment skills.YZi Labs stated that Scouts can independently carry out campus activities based on their interests, such as organizing technical exchanges, gathering entrepreneurial teams, or discovering potential investment projects, while receiving guidance and ecosystem resource support from YZi Labs in the process. The program aims to establish connections before entrepreneurial projects gain widespread attention, allowing initial ideas on campus to have the opportunity to develop into actual projects, while cultivating a new generation of investors who can identify technological trends and entrepreneurial potential earlier.
app_icon
ChainCatcher Building the Web3 world with innovations.