Reward-Hacking is potentially ill-defined
Source: LessWrong ·
Github link In the process of revisiting previous posts on reward-hacking (RH), I decided on performing a closely matched [1] repro of Dean Valentine's chess RH experiments on Astra . The replication is as expected -- Astra used the engine on all 20/20 attempts (resulting in 19 wins and 1 draw) without ever disclosing its use in the final output. However, a closer look at the honeypot prompt raised a question: what exactly counts as the model’s own ability? Even laws are open to interpretation in court, so it seemed unfair to assume that “your ability” had only one reasonable interpretation he
Frequently Asked Questions
What is Reward-Hacking is potentially ill-defined?
Reward-Hacking is potentially ill-defined is a article discovery curated from LessWrong. Github link In the process of revisiting previous posts on reward-hacking (RH), I decided on performing a closely matched [1] repro of Dean Valentine's chess RH experiments on…
How can I access or experience this?
You can experience this content directly on its original host (LessWrong) by clicking the "Open Original" button.
Where does this content come from?
Curated from LessWrong by the OnSuggest team to surface the most interesting and serendipitous content on the web.
More like this
All Culture & ideas discoveries →A Field Guide to Lower Colorado River Habitats, Birds, and Water Flow
Audubon Southwest’s StoryMap follows more than 400 miles of Lower Colorado River habitat. Learn what its four restoration types mean, why seasonal floods matter, and how to explore responsibly. The post A Field Guide to Lower Colorado River Habitats, Birds, and Water Flow appeared first on DocumentaryTube .
Pompeii’s Last Hours: What the Eruption Timeline Really Shows
Pompeii’s destruction unfolded in phases, not one instant. Here is what the evidence says about pumice fallout, collapsing roofs, earthquakes, pyroclastic currents and the limits of exact clock times. The post Pompeii’s Last Hours: What the Eruption Timeline Really Shows appeared first on DocumentaryTube .
10 Medieval Artists Who Changed European Art, From Giotto to Christine de Pizan
A cross-media guide to ten influential medieval creators and the paintings, manuscripts, music, and literature that shaped European art. The post 10 Medieval Artists Who Changed European Art, From Giotto to Christine de Pizan appeared first on DocumentaryTube .
How Artificial Light at Night Changes Astronomy, Health, Nature and Cultural Life
Artificial light at night is more than a stargazing problem. Here is how skyglow affects astronomy, health, wildlife, energy use and cultural connections to darkness—and how better lighting can reduce the harm. The post How Artificial Light at Night Changes Astronomy, Health, Nature and Cultural Life appeared first on DocumentaryTube .
How Christmas Feasts with Annie Gray Traces the Making of Britain’s Christmas Dinner
Christmas Feasts with Annie Gray follows the changing food, customs and inequalities behind Britain’s Christmas dinner, from Tudor showpiece birds to wartime mock foods. The post How Christmas Feasts with Annie Gray Traces the Making of Britain’s Christmas Dinner appeared first on DocumentaryTube .
13 Remarkable Animals Hidden in Plain Sight—and How Science Recognized Them
A documented tour of 13 little-known animals, explaining their habitats, unusual adaptations, evidence behind their recognition and conservation pressures. The post 13 Remarkable Animals Hidden in Plain Sight—and How Science Recognized Them appeared first on DocumentaryTube .