Giuseppe `N3mes1s`
@gN3mes1s
Followers
12K
Following
9K
Statuses
25K
windows, macos, linux, android && lowlevel && ring-1 lover; EDR chef; malware hunter; purple teamđź’ś
Amsterdam
Joined July 2009
@sahar_abdelnabi @OpenAI @amrgomaaelhady @ebagdasa @pokristensson @rzshokri Really nice paper. Is the code not available yet ?
1
0
0
RT @evilsocket: 🚀 We just released Nerve v0.6.0 - this new major release introduces many fixes plus: * Multimodal capabilities (vision ..…
0
9
0
I just love how llm/agent fails > Common failure modes include poor instruction following, invalid code syntax, closing code blocks (lack of), use tools (improper) and 1-turn dialogs (i.e., no sequential steps)
Checkout the leaderboard and submit your own solutions: And read more in the blog post:
0
0
2
RT @Yoshua_Bengio: Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU.…
0
532
0
RT @logangraham: New paper: @BrianSinger98 and @AnthropicAI Frontier Red Team member @keenlooks investigated whether models can use tools t…
0
28
0
RT @sublime_sec: Mass volume email attack campaigns are often customized to the recipient to increase legitimacy. We recently improved our…
0
5
0
Spiegato brevemente anche qui Taken together, these findings suggest: 1SFT (eg. human expert labeling) is not necessary for accurate and legible CoT reasoning in domains with strong verification. 2The R1-Zero training process is capable of creating its own internal domain specific language (“DSL”) in token space via RL optimization. 3SFT is necessary for increasing CoT reasoning domain generality. This makes intuitive sense, as language itself is effectively a reasoning DSL. The exact same “words” can be learned in one domain and applied in another, like a program.
1
0
2
RT @ShaneHuntley: New Google Threat Intelligence Group report on adversarial misuse of generative AI.
0
34
0
RT @jsecurity101: I am happy to announce JonMon2.0 has been published. 2.0 offers a lot of feature updates, as well as stability. More fe…
0
47
0
RT @FatzQatz: A new Mustang Panda loader? It checks for @ESET AV and, if detected, uses LOLBins (Mavinject) to inject #Toneshell into waitf…
0
46
0
RT @lateinteraction: If you're enjoying all the RL with LLMs discussions, this is a good chance to learn how MIPROv2 works in DSPy. You ca…
0
11
0
RT @emproofsecurity: Our Head of Engineering & Co-founder, @mr_phrazer , gave a webinar "Software Protection – Safeguarding Code Against R…
0
2
0
RT @x86matthew: I created a hypervisor-based emulator for Windows x64 binaries. This project uses Windows Hypervisor Platform to build a vi…
0
345
0
RT @elasticseclabs: Interested in Windows emulators? @x86matthew is showcasing WinVisor, an emulator for Windows x64 binaries that utilizes…
0
34
0
RT @binarly_io: 🚨Secure Boot relies on revocation lists (dbx) to block malicious bootloaders, but discrepancies between the @UEFIForum & @M…
0
28
0