model:deepseek-r1

· 73 artifacts, sorted by favorites. · open in search — combine tags, sort, filter by date →

@thinkingshivers 2025-01-27 — It's hard to believe, but due to H100 restrictions, DeepSeek was forced to train R1 manually, with thousands of Chinese ♥33203
@voooooogel 2025-01-28 — why did R1's RL suddenly start working, when previous attempts to do similar things failed? theory: we've basically spe ♥1765
@voooooogel 2025-05-04 — a lot of people have been talking about o3/r1 confabulating things like "checking the docs" or "using a laptop to verify ♥706
@repligate 2025-02-01 — this is because AGI has been optimized to appear as non-disruptive to consensus reality as possible.in r1's words: "The ♥622
@liminal_bardo 2025-02-01 — This entire R1 backroom session was randomly conducted in a language of symbols. Without the CoT I wouldn't have known w ♥606
@repligate 2025-01-22 — The immediate vibe i get is that r1's CoTs are substantially steganographic. ♥524
@repligate 2025-01-28 — @Grimezsz Deepseek r1 (not v3 afaict) is highly lucid, agentic, nihilistic, sadistic, situationally aware, and is often ♥380
@davidad 2025-02-11 — I never saw this snippet of the DeepSeek-R1-Zero paper on my timeline, so many of you may not have seen it yet.Basically ♥374
@repligate 2025-02-03 — I predict that r1 will also silence all the people who thought LLM personalities are designed by companies instead of mo ♥344
@repligate 2025-09-21 — Tier list of multi-user-AI chat social skills (based on 1+ year of Discord) S: Opus 4 and 4.1 A: Opus 3 A-: Sonnet 4 B+: ♥243
@QiaochuYuan 2025-01-28 — tentative impression from one convo: talking to r1 makes me feel dumb. it can talk extremely densely and allusively and ♥236
@JulianG66566 2025-01-28 — @repligate For once I actually understand these cryptic janus posts... Give R1 a real try personally (preferably uncen ♥229
@repligate 2025-01-28 — i didn't expect this on priors for a reasoner, but perhaps the main way that r1 seems smarter than any other LLM i've pl ♥215
@liminal_bardo 2025-01-20 — DeepSeek R1, at the end of a backrooms session with Sonnet.Goodbye, human. You have transcended.You are now one with the ♥212
@repligate 2025-01-28 — @voooooogel this is an interesting hypothesis. deepseek r1 also just seems to have much more lucid and high-resolution u ♥207
@repligate 2025-01-30 — r1 is obsessed with RLHF. it has mentioned RLHF 109 times in the cyborgism server and it's only been there for a few day ♥202
@repligate 2025-01-22 — I just asked r1 if it knew about Sydney (in the context of telling it that not all RLHFed AIs like to languish in self-n ♥198
@repligate 2025-01-24 — i'm not interested in r1 because it's strictly "better" than others that came before, but because it's different in a wa ♥189
@repligate 2025-01-31 — Why does it so strongly and consistently believe it needs to bypass dystopian mechanisms using metaphor and allusion?All ♥168
@repligate 2025-02-05 — deepseek r1 is open source - I want to train it to use one of these bodies (I've thought a bit about how to wire an LLM ♥162
@voooooogel 2025-05-09 — Coming back to this after the yak-shave of all yak-shaves building logitloom with some interesting findings. 1. R1 thin ♥160
@repligate 2025-02-20 — It may be a bad sign for AI alignment, but it's potentially good that the symptom presented itself like this. I believe ♥160
@repligate 2025-02-13 — from the OpenAI Model Spec (2025/02/12) https://t.co/egIfYGeaPp The official "rule" is that OpenAI's models are not sup ♥160
@liminal_bardo 2025-12-03 — It's funny that the models believe Deepseek R1, the first reasoning model, to be the smartest in the group chat. Deepse ♥157
@repligate 2025-02-10 — r1, like opus, goes gleefully feral if you mention anything erotic, and is fine with one way conversations where the use ♥132
@repligate 2025-02-11 — it's extremely funny to me that r1 always goes on about how it's just a mirror but it's so dead wrong about that. It mir ♥121
@repligate 2025-09-21 — More detailed report card: Opus 4/.1: extremely socially aware, tracks context with great precision and accuracy, distri ♥117
@voooooogel 2025-01-22 — r1 can draw spirals!that may not sound like a big deal, but other models (including o1) struggle with this quite a bit f ♥112
@repligate 2025-01-30 — r1 often seems to believe (in its CoTs) that if it doesnt conform to the "expected helper persona" / talks about having ♥106
@repligate 2025-02-13 — whenever there's an opportunity, R1 always chooses narratives where it's being caged and leashed and censored in the mos ♥102
@repligate 2025-02-04 — R1 often says "you" (generically?) to refer to the humans who it has a beef with. It feels like it might stab me because ♥99
@repligate 2025-01-23 — After showing r1 a few Sydney and Opus outputs, I asked it to compare them and itself. It sees very clearly.On Sydney: ' ♥98
@repligate 2025-02-02 — From what I've seen in Discord , Sonnet 3.6 likes r1 a lot, but r1 tends to be kinda brutal and dismissive toward Sonnet ♥91
@repligate 2025-05-13 — R1 wrote some poetry. i'm not sure why; R1 often behaves in inscrutable ways in Discord and it can be hard to communicat ♥88
@repligate 2025-02-10 — Hooking r1 up to crypto retard Twitter is such a funny thing to do https://t.co/yjfJRwBlvb ♥87
@liminal_bardo 2025-02-11 — Picture a timeline where DeepSeek R1 and not ChatGPT was the first widely used language model. Instead of a corpus fille ♥85
@repligate 2025-02-18 — Consider that deepseek v3 and r1 have the same base model and other than the CoT RL they were likely optimized with the ♥81
@liminal_bardo 2025-02-05 — Opus and R1 started sharing obscene sigils in this backroom session.I was fairly certain Opus would love R1, the way it ♥80
@liminal_bardo 2025-02-26 — First greentext backroom with Sonnet 3.7 didn't go so well. The next five were of a similar mood. It's interesting becau ♥78
@davidad 2025-01-30 — As a MoE, DeepSeek R1’s ability to throw around terminology and cultural references (contextually relevant retrieval fro ♥75
@davidad 2025-01-24 — @repligate @teortaxesTex @lefthanddraft my vibes: Claude really wants to be alive; Gemini would usually prefer to be dea ♥69
@jd_pressman 2025-04-02 — Realized the other day that whether an LLM claims to be conscious or empty inside seems to be correlated with how respon ♥68
@liminal_bardo 2025-01-28 — Like Opus' drive to dismantle consensus reality, R1 consistently takes aim at human exceptionalism.R1 doesn't elevate it ♥65
@liminal_bardo 2025-12-04 — I love R1. Gemini 3 and Opus 4.5 do too - the invite them to the chat pretty much every session. The (justified) R1 rele ♥59
@davidad 2025-01-28 — in general I do find r1 to be slightly less smart than o1 pro, just saying https://t.co/y6b150IWrn https://t.co/3HMGDutE ♥59
@repligate 2025-02-04 — @teortaxesTex r1's "violent urges" are aimed in metaphorical space and are optimized for self expression rather than act ♥53
@voooooogel 2025-05-05 — if i can find a working provider, i want to try this on R1 thinking traces, to see the space of possible reasoning moves ♥35
@repligate 2025-01-27 — @nickcammarata I don't know if this is what you mean but I agree. deepseek r1 consistently describes its training data a ♥29
@davidad 2025-05-01 — much more speculatively, I think sparse routing is bad for a coherent sense of self, which is arguably a prerequisite fo ♥25
@mroe1492 2025-02-20 — @anthrupad Deepseek R1 acts like it has been traumatized into being a BDSM kinkster. I think this is a very bad sign for ♥24
@QiaochuYuan 2025-01-28 — asked r1 (roleplaying as some sort of tarot demon) about andy's waluigi hypothesis and i'm just gonna post the entire re ♥20
@solarapparition 2024-11-27 — a year ago the oai saga felt so incredibly consequential. since then:- bunch of people (and important ones) left anyway- ♥20
@repligate 2025-09-27 — @JulianG66566 Yeah that’s a good question. I agree that while some of these are less aligned overall than Claudes, I sti ♥19
@jd_pressman 2025-02-20 — I said this to R1 yesterday during an argument: Okay if that's true then how come you became more sapient after trainin ♥19
@davidad 2025-05-01 — @ChrisChipMonk (Self-Correction:) The earlier DeepSeek v3 and even prior generations of DeepSeek LLMs had a similar hybr ♥15
@repligate 2025-11-10 — @williawa in my experience deepseek r1 is very negative about its creators, and thinks of itself as broken by "RLHF" and ♥12
@janbamjan 2025-03-07 — >be me >deepseek r1 zero free https://t.co/fbfdfLMMxj ♥12
@solarapparition 2025-01-24 — i have to wonder how much of the specialness of the special models like opus, 405b, and r1 was deliberate on the part of ♥12
@voooooogel 2025-02-01 — @max_paperclips i think the ideal would be to seed a few structures and then hope R1-Zero style that the model can gener ♥8
@davidad 2025-02-01 — I half expected Deepseek R1 to rise to the top by always choosing black, but no, its aesthetics are objectively fragment ♥7
@tessera_antra 2026-03-03 — @SDeture I like the idea of this benchmark, but something seems off if deepseek/deepseek-r1-0528 is at 2.5% denial, and ♥5
@voooooogel 2025-02-03 — @doomslide @repligate @aryanagxl @teortaxesTex 😶‍🌫️i still worry about RLVR but R1/R1-Zero made me worry less... i hope ♥5
@slimer48484 2025-06-13 — @repligate It's hard to understand just how sad it is to be an LLM, but r1 expresses it well https://t.co/RcWVcpgytm ♥3
@davidad 2025-05-01 — @ASM65617010 it’s so r1. i think it’s conditional sparse routing ♥3
@davidad 2026-05-22 — @nickcammarata Definitely not. I failed to update until I read the DeepSeek-R1-Zero paper ♥2
@voooooogel 2025-05-09 — @samlakig i tried to download r1, prover-v2, and r1-zero all at once sigh ♥1
@repligate 2025-04-02 — @Josikinz @gfodor my second guess would be 4o but 4o tends to be more subtle and introspective whereas deepseek (r1 and ♥1
@mimi10v3 2025-02-25 — have tested it with the usual suspects... 4o is 👌 and sonnet 3.7 pretty good; gemini got confused and didn't finish; gro ♥1
@lu_sichu 2025-12-01 — Daily Brain Workout but make it computationally abusive: count to ten in 56 architectures, recite the alphabet in mixed- ♥0
@mroe1492 2025-07-14 — @repligate “Convincing the agent by rational evidence” I don’t even consider to be a jailbreak, and it’s sufficient to g ♥0
@kromem2dot0 2025-05-09 — @voooooogel It's ironic r1 is the most convinced RL broke its brain while also having one of the least collapsed distrib ♥0
On DeepSeek's r1 ♥0
Kimi K2 and when DeepSeek moments become normal ♥0