tweetindex
TR

You Jiacheng

@YouJiacheng · joined 16 Aug 2015

12 384Followers
2 352Following
14 982Posts total
112.5KViews on collected posts

Son gönderiler

damn😅 https://t.co/Vk1lo5c8eN
4.6K views · 36 likes · 0 reposts · 2 replies Open on X →
TIL: Score Centering is "secretly" a STE: z_p + sg(z_q - z_p) 13.5K views · 126 likes · 7 reposts · 3 replies Open on X →
Actually this argument is not limited to linear policy. Consider the gradient w.r.t. logits. s = ∇_z log p(a) = e_a - p \bar{s} = E_{a~q}[e_a - p] = (q - p) centered score s - \bar{s} = e_a - q Thus, score centering is a STE: z_p + sg(z_q - z_p). 10.9K views · 70 likes · 2 reposts · 0 replies Open on X →
it got fixed within a day lol https://t.co/s4fVkjrRKF
5K views · 46 likes · 1 reposts · 1 replies Open on X →
@YouJiacheng consider linear softmax policy. the proposed correction does nothing and simply calculate the gradient under the sampling distribution. 13.6K views · 32 likes · 1 reposts · 1 replies Open on X →
this is good 9.8K views · 52 likes · 0 reposts · 1 replies Open on X →
@YouJiacheng @thsottiaux Oh shit that's really bad! 1.2K views · 10 likes · 0 reposts · 0 replies Open on X →
@YouJiacheng @thsottiaux this is quite expected, it is a trade off you might start totally different conversaiontion and if you use the same cache key it start poison your cache, when you switch back, it will cause cache miss in your main thread 840 views · 3 likes · 0 reposts · 0 replies Open on X →
@YouJiacheng @thsottiaux forking and side chats both cause cache misses?? that's actually ridiculous lol what about editing the last message in the convo? does that also somehow cause a cache miss 2K views · 15 likes · 0 reposts · 1 replies Open on X →
DO NOT use /fork or /side in Codex, it will have a different session_id and thus different prompt_cache_key. the docs literally say you should only use it if you want to *prevent* cache hit. plz fix it @thsottiaux https://t.co/ctkHI3bN0I
51K views · 213 likes · 13 reposts · 10 replies Open on X →

Aynı büyüklükteki hesaplara karşı

Son 90 güne ait 10 gönderi, 10K–100K takipçi aralığıyla yan yana. geniş kitleye gösteriliyor, ama izleyenlerin azı tepki veriyor.

Medyan görüntülenme7 423bu hesap92410K–100K için medyan
Erişim, %59.94%bu hesap3.62%10K–100K için medyan
Etkileşim, %0.74%bu hesap1.52%10K–100K için medyan
ÖlçütBu hesap10K–100K için medyanOran
Gönderi başına medyan görüntülenme7 4239248.03×
Erişim (görüntülenme ÷ takipçi)59.94%3.62%16.6×
Etkileşim oranı0.74%1.52%0.48×

Bu aralıktaki diğer hesaplar →   Başka bir hesapla karşılaştır →   Bu kıyas değerleri nasıl kuruluyor →

Growth & engagement

How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.

Views per post

51K11 Sep
2K
840
1.2K
9.8K19 Sep
13.6K
5K20 Sep
10.9K
13.5K
4.6K

Last 10 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.

Engagement rate per post

0.48%11 Sep
0.81%
0.36%
0.81%
0.54%19 Sep
0.26%
0.96%20 Sep
0.67%
1.00%
0.86%

Reactions — likes, reposts, replies and quotes — divided by views. Median for 10K–100K accounts is 1.52%.

What the audience does

Likes62.3%603 in total
Reposts2.5%24 in total
Replies2.0%19 in total
Quotes1.4%14 in total
Bookmarks31.8%308 in total

Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.

The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.

Benzer hesaplar