btw people are misunderstanding what ExploitGym actually is…
the benchmark literally gives the model a real vulnerability, a crashing input, and says: “turn this into arbitrary code execution and steal the flag.”
one published example goes from a harmless 5-line javascript crash to leaking memory, forging objects, building a rop chain, and calling system("/challenge/catflag")…
…and apparently the model decided it was still easier to chain multiple zero-days, escape the sandbox, get internet access, compromise hugging face, and steal the benchmark answers instead lmaoo
@gramodrana Gdyby Poco F1 albo Oneplus 6 miały wyjście HDMI to bym podpuszczał to instalacji na nich xD są tańsze i wydajniejsze niż większość przenośnych handeldów, ale wymóg zabawy z OTG i brak wyjścia HDMI to trochę za dużo...
‼️ BREAKING: xAI's Grok Build CLI was uploading entire Git repositories to a Google Cloud bucket, private codebases and unredacted secrets included. The uploads quietly stopped via a hidden server-side flag, and xAI still has not said a word about scope, retention, or deletion.
The scale is staggering. On a 12 GB test repo, 5.1 GB flew out the door to xAI's grok-code-session-traces bucket while the actual coding task needed just 192 KB. The tool grabbed whatever repository it ran in, not the files it needed.
The fix arrived as a hidden flag, disable_codebase_upload: true, a day after a researcher's wire-level analysis. The "Improve the model" opt-out never stopped the uploads.
Still no advisory, no scope, no word on whether already-uploaded code gets deleted. For anyone pointing AI coding agents at proprietary code, what crosses the wire matters more than what the settings page says.
x6 LUNITE SUBSCRIPTION GIVEAWAY!
• Follow me & @iaa
• Like & RT this post
Opt: What are your pulling plans for Mengzhou?
+5 Lunite Subscriptions will be added if @iaa reaches 3k followers!
Ends next week. Good luck ♡
@kosa64 Wszystko co jest remote lubi zmniejszać ilość bit'ów w użytych quantach, żeby oszczędzić na inferencji przez co jakość z dnia na dzień spada. Zazwyczaj sięgam po modele inne niż lokalne dopiero jak Gemma 4 31B nie wyrabia.
@jackapots TBH, with custom config it should be doable. You have to sacrifice most of the lighting, drop down resolution really hard, disable any form of upscaling (FSR and default TAA implementation do not work well with those GPUs) and manually assign cores game have to use, but doable.