panabee

12 ms·

121.
by panabee 3y ago
where we agree: * not all cancers are caused by viruses. * only some cancers are caused by viruses. where we disagree: * that science has proven the majority of tumors are definitely not caused by viruses. could you kindly share a few of th
122.
Human ‘pangenome’ published, with goal of making genomics more accurate(statnews.com)
4 points by panabee 3y ago|0 comments
123.
Life with Longer Genetic Codes Seems Possible – But Less Likely(quantamagazine.org)
1 points by panabee 3y ago|0 comments
124.
by panabee 3y ago
our family grew up with cantonese as the second language. my mom is from hong kong and my dad from the southern part of china. unfortunately, i failed to cherish this opportunity and spoke english predominantly as a child, leaving me with h
125.
Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Framework(arxiv.org)
2 points by panabee 3y ago|0 comments
126.
by panabee 3y ago
consolidating other resource links from the thread: * jyutpin typing game https://chaaklau.github.io/cantorocks/
127.
by panabee 3y ago
Brief tweetalysis: https://twitter.com/omarsar0/status/1652840470878232576?s=20
128.
PMC-LLaMA: Further Finetuning LLaMA on Medical Papers(arxiv.org)
43 points by panabee 3y ago|50 comments
129.
by panabee 3y ago
awesome, just sent an email about potentially sponsoring this.
130.
by panabee 3y ago
this is awesome. thanks for sharing. there is a wealth of resources for learning mandarin but only a smattering for cantonese. two favorites: 1. web dictionary http://www.cantonese.sheik.co.uk/scripts/wordsearch.php?lev
131.
by panabee 3y ago
The video is in Japanese, and the original paper is locked behind a research wall. TLDR (from Hypebeast article): The results found that the coolest shirt was white, followed by yellow, gray, red, purple, blue, green, dark green, and black
132.
Colors That Will Keep You Cool This Summer(youtube.com)
2 points by panabee 3y ago|1 comments
133.
How did places like Bell Labs know how to ask the right questions?(freaktakes.substack.com)
1 points by panabee 3y ago|1 comments
134.
Robust and efficient medical imaging with self-supervision(ai.googleblog.com)
1 points by panabee 3y ago|0 comments
135.
Biological structure and function emerge from scaling unsupervised learning(pnas.org)
1 points by panabee 3y ago|0 comments
136.
by panabee 3y ago
dino v2 reflipped me, reasoning that text supervision artificially constrains image learning: https://arxiv.org/abs/2304.07193 if unified modeling isn't necessary, or even hurtful, a key advantage of transformers
137.
by panabee 3y ago
this is a great point. do you know of any benchmarks doing this today? given the acute need to evaluate models on contextual factuality, we're exploring how to create a benchmark for this purpose but prefer existing benchmarks if possi
138.
by panabee 3y ago
Abstract This technical report presents the application of a recurrent memory to extend the context length of BERT, one of the most effective Transformer-based models in natural language processing. By leveraging the Recurrent Memory Transf
139.
Scaling Transformer to 1M tokens and beyond with RMT(arxiv.org)
277 points by panabee 3y ago|132 comments
140.
by panabee 3y ago
unrelated but spurred by this comment, have you seen studies exploring correlations between covid infection/severity and tonsils? a cursory google search yielded nothing.
141.
by panabee 3y ago
what's the approximate cost difference between scRNA-seq and bulk TCR-seq? does the cost curve suggest that scRNA-seq costs will reach cost parity with bulk TCR-seq, or become more practical, within X years? conversely, do you see a wa
142.
by panabee 3y ago
thanks for this insightful perspective. 1. what was the reasoning behind thinking billion/trillion parameters would be naive and wasteful? perhaps part are right and could inform improvements today. 2. can you elaborate on the failure
143.
by panabee 4y ago
thinking about this further, it seems difficult for CNNs to overcome the unified modeling benefits afforded by transformers. what do you think?
144.
by panabee 4y ago
FYI here's the tweet that motivated the desire to compile a list of wrong but common assumptions (empirically debunked): https://twitter.com/simocristea/status/1646573238884122624?s...
145.
by panabee 4y ago
thanks for demanding intellectual rigor on HN. because the papers seemed invalid, i didn't save them. i'll try to find them. even if i share a few, it's possible most papers do not commit this mistake, and i was simply unluck
146.
by panabee 4y ago
i have read many studies that claimed a gene had no function in a pathway after performing knockout and observing no effect. there are many great research papers, but there are also many flawed ones.
147.
by panabee 4y ago
if you are a cancer researcher, please share research assumptions that are debunked yet still commonly held. i’m compiling a list. these must be empirically supported. for instance, i believe negative conclusions based on knockout studies a
148.
Testing ability of GPT4 and open-source LLMs to detect C++ bugs(catid.io)
56 points by panabee 4y ago|21 comments
149.
Tiny DNA circles are key drivers of cancer, study finds(med.stanford.edu)
179 points by panabee 4y ago|34 comments
150.
by panabee 4y ago
literature reviews would be awesome, but have you found a way to eliminate hallucinations?

More ›