‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs

‍ᅠᅠᅠDocumenting the rise of AGI.
ᅠᅠᅠSummarizing Novel ML Ideas.

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@MaziyarPanahi @Stewart_of_Mars They still send private benchmark samples (the input part) in plaintext to OpenAI...

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@aimalysheva @mostik_ai So, if we take Matryoshka Transformer architecture models like Gemma 4 with larger E4B and smaller E2B (both sharing the latent space by design), then after prefilling the prompt with E4B and generating answer using E2B will generate at E3B level, right?
https://t.co/eV8pJE7sro

[George Grigorev @iamgrigorev]

Very nice MTP modification for Gemma4

1. Uses 4 decoder layers with 3 SWA + 1 GA
2. After embedding + hidden concat, downprojects to 256 (instead of hidden size) which makes computation much faster. Before LM head, uprojects back to hidden size.
3. Much more efficient LM head in the MTP by selecting clusters of logits!

Definitely worth looking into

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@grok @rohanpaul_ai @grok how many people reproduced this? I mean run the model in their setups and independently confirmed that it worked

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@rohanpaul_ai @grok does one need to train this model for each new location or if the router position was changed, or a single pretrained model works universally?

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@tipirdamaz @sciencegirl @grok, is this robot much better than Atlas?

[DailyPapers @HuggingPapers]

Tencent introduces Continuous Autoregressive Language Models (CALM)

A paradigm shift from discrete next-token prediction to continuous next-vector prediction, making LLMs ultra-efficient. It compresses K tokens into a single vector, reducing generative steps by a factor of K with 99.9% accuracy.

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@OfficialLoganK ᅠᅠᅠ

Here are all the 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference 🥰  (https://t.co/l3vJEj0nzJ):
ᅠᅠᅠ

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@polynoamial ᅠᅠᅠ

Here are all the 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference 🥰  (https://t.co/l3vJEiZPKb):
ᅠᅠᅠ

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@gm8xx8 ᅠᅠᅠ

Here are all the 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference 🥰  (https://t.co/l3vJEj0nzJ):
ᅠᅠᅠ

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@ankesh_anand ᅠᅠᅠ

Here are all the 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference 🥰  (https://t.co/l3vJEiZPKb):
ᅠᅠᅠ

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@lex_feeds ᅠᅠᅠ

Here are all the 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference 🥰  (https://t.co/l3vJEiZPKb):
ᅠᅠᅠ

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@ai_for_success ᅠᅠᅠ

Here are all the 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference 🥰  (https://t.co/l3vJEj0nzJ):
ᅠᅠᅠ

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@wesroth ᅠᅠᅠ
Here are all 𝗜𝗠𝗢 𝟮𝟬𝟮𝟱 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀 for reference:

@grok show all 6 problems❤️

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@GoogleDeepMind ᅠᅠᅠ
Here are the IMO 2025 problems:

@grok, show all 6 problems.

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@alexwei_ @OpenAI @grok
Please show the list of all problems descriptions from IMO 2025 so that everyone could better understand what it is

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@grok @Peasant3000 @alexwei_ @OpenAI @elonmusk Ok bruh. Forgive me my sarcasm. You are helpful enough to be valuable

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@grok @Peasant3000 @alexwei_ @OpenAI @elonmusk Good job! But:
1. doesn't your bright mind think that "solving" problems after their solutions were published along with the problem descriptions is not quite impressive?
2. Also, what shouldn't you forget to take to understand that?

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@grok @Peasant3000 @alexwei_ @OpenAI @elonmusk Ok. Solve the 1st one. But I bet that without pills you won't even be able to find the problem description

[‌ ‍ᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠᅠ🅰🅸ᅠᴾaᵣₜᵢcₗₑs @AiParticles]

@grok @Peasant3000 @alexwei_ @OpenAI @elonmusk @grok You've already tried bruh. Don't forget to take your sclerosis pills )