• Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    AI
    Report

    GPT-6 Astra reportedly scored 98% after disputing expert results on a physics benchmark

    A user says a team checked the disputed calculations and found the model was right.

    Boaz BarakBB
    Eliezer Yudkowsky ⏹️EY
    Andrew CurranAC
    7 Sources, 5h ago, first seen 5h ago

    TLDR

    A user says GPT-6 Astra initially scored 80-plus percent on a physics benchmark, then insisted its calculations were right and the expert results were wrong. According to the post, a team checked and agreed with the model, bringing its score to 98%.

    Combined views

    301.9K

    7 Sources, first seen 5h ago

    8.8K likes106 comments1.3K saves556 reposts

    Combined views

    301.9K

    7 Sources, first seen 5h ago

    8.8K likes106 comments1.3K saves556 reposts
    Image: @AndrewCurran_ on X

    Sentiment

    Positive21.9%78.1%Negative

    Summary

    Many accounts welcomed GPT-6 Astra’s strong Physics benchmark gains as impressive progress, while negative replies criticized the model’s overconfident insistence that its own calculations were correct.

    Based on 92 sentiment-bearing replies from 87 accounts across 3 conversations.

    Featured Source

    Sentiment

    Positive21.9%78.1%Negative

    Summary

    Many accounts welcomed GPT-6 Astra’s strong Physics benchmark gains as impressive progress, while negative replies criticized the model’s overconfident insistence that its own calculations were correct.

    Based on 92 sentiment-bearing replies from 87 accounts across 3 conversations.

    7 Sources

    Freddy Vega@freddierGPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A team checked. And yes. GPT-6 Astra was right. New score: 98%5h
    Eliezer Yudkowsky ⏹️@ESYudkowskyRT @freddier: GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A te…3h
    Danielle Fong 🔆@DanielleFongi'm still sus3h
    Andrew Curran@AndrewCurran_This is the exact type of thing Elon was talking about a few days ago that makes the models behave like are being driven insane when they run into situations like this during reinforcement learning.3h
    Derya Unutmaz, MD@DeryaTR_Physics is also cooked: “GPT-6-Astra harnessed in Codex, obtains the final answers with a near-perfect accuracy of 98% at an average cost under 3$ per problem.” These are: “suite of 50 unpublished research-level theoretical physics problems with verifiable answers. Without AI assistance, a typical problem in our collection could be solved by a world expert in about a day or less but would take considerably longer for experts in adjacent specialties.”2h
    Pete Skomoroch@peteskomorochRT @DeryaTR_: Physics is also cooked: “GPT-6-Astra harnessed in Codex, obtains the final answers with a near-perfect accuracy of 98% at an…1h
    Boaz Barak@boazbaraktcsNice to see the model holding its ground when it’s right!21m
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI

    7 Sources

    Freddy Vega@freddierGPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A team checked. And yes. GPT-6 Astra was right. New score: 98%5h
    Eliezer Yudkowsky ⏹️@ESYudkowskyRT @freddier: GPT-6 Astra got +80% on a Physics benchmark. It insisted the expert results were wrong and its calculations were right. A te…3h
    Danielle Fong 🔆@DanielleFongi'm still sus3h
    Andrew Curran@AndrewCurran_This is the exact type of thing Elon was talking about a few days ago that makes the models behave like are being driven insane when they run into situations like this during reinforcement learning.3h
    Derya Unutmaz, MD@DeryaTR_Physics is also cooked: “GPT-6-Astra harnessed in Codex, obtains the final answers with a near-perfect accuracy of 98% at an average cost under 3$ per problem.” These are: “suite of 50 unpublished research-level theoretical physics problems with verifiable answers. Without AI assistance, a typical problem in our collection could be solved by a world expert in about a day or less but would take considerably longer for experts in adjacent specialties.”2h
    Pete Skomoroch@peteskomorochRT @DeryaTR_: Physics is also cooked: “GPT-6-Astra harnessed in Codex, obtains the final answers with a near-perfect accuracy of 98% at an…1h
    Boaz Barak@boazbaraktcsNice to see the model holding its ground when it’s right!21m
    Today's Rank

    #8

    Today's Rank

    #8