Maya Moritz
@mayarmoritzCriminology PhD Student @Penn. Keen to tell you about place-based crime intiatives, bouldering blunders, & my @duolingo streak. Here to learn. https://t.co/yTO4jKBZFM
Similar User
@EconUniMannheim
@Dalya_Bk
@IE_Castillo
@PavlovicVal
@TheSandyXie
@ScreenPhillip
@Candido_Leonard
@J0nasCasper
@Nicoledaltroy
So delighted to see this Omnia article out about my current project on murals and crime + the general struggle of data collection!
Criminology PhD student Maya Moritz is collecting data to see if murals affect Philadelphia's crime rate. Her findings suggest that crime does decrease in neighborhoods after a mural goes up, and the effect can last well into the future. bit.ly/3JjzIEX @Penn
Another great day at @emnlpmeeting! Research shared by @johnamqdang, @ahmetustun89, and @d_aumiller, and our small 🐐 friend is exploring the sites of Miami 🌴. Our team will be at the @cohere booth for the rest of the day. Stop by to say hello.
Waiting in line at the conference social right now, and browsing Twitter to see what you should check out tomorrow at #EMNLP2024? Stop by our poster in Riverfront Hall tomorrow morning between 10:30-12, to hear more from @ahmetustun89 and me!
How does quantization affect multilingual LLMs? 🌍 For wide adoption, multilingual LLMs must be highly-performant *and* lightweight. 📈 🪶 We analyze SOTA multilingual LLMs in 23 languages under various quantization techniques to find out! 📜 arxiv.org/pdf/2407.03211
 Welcome to the webinar: AI, machine learning and related new technologies in Criminology  Thursday 7th of November 17:30 CET / 11:30 EST / 08:30 PST Register Here: bit.ly/3Uh6Naw @ASC_DCP @YuhaoKang @sakraplatser_n @GabrielGliori @Riley_NEU @mayarmoritz
🚨 New Paper Alert 🚨 SO hyped about this work led by Jonny - he did an amazing job during his internship. It's the first paper about what we think are key challenges when evaluating general-purpose LLMs! Check the paper here: arxiv.org/abs/2410.03608
What if improving LLM evaluation and generation was as simple as using a checklist? Introducing TICK ✅ (Targeted Instruct-evaluation with ChecKlists) and STICK 🏒 (Self-TICK) Work done @cohere with supervision from @_rockt, @j_foerst, @d_aumiller & @W4ngatang 1/n
I've spent all of five minutes coming up with these categories, fight me.
Mural Arts unveils first of six youth anti-violence mural project The state crime and delinquency agency funded a new mural in Fitler Square. It is the first of six murals designed by children. whyy.org/articles/mural… Cc @mayarmoritz
1/ New paper alert! 🚨 Leakage of popular datasets is always assumed, but rarely checked. In this work, we provide a *refreshed dataset* in the style of HumanEval/MBPP for evaluating LLMs on code generation tasks: lbpp Use the dataset today: huggingface.co/datasets/Coher…
"HumanEval leaked". You probably heard that one already. But what is the evidence? In this preprint we look into 3 hypothesis: direct leakage, indirect through synths & overfitting to eval Key to our findings is a fresh test set we are also releasing! arxiv.org/abs/2407.07565
Super interesting analysis! And now I know what Sankey diagrams are.
After hearing conflicting info about the #lawschool prof market, I decided to use the wonderful data from @sarahlawsky's blogposts on entry-level hiring to explore the PhD/JD market: mayamoritz.wordpress.com/2024/07/05/blo… A few take-aways below:
United States Trends
- 1. Ravens 76,1 B posts
- 2. Steelers 103 B posts
- 3. Bears 110 B posts
- 4. Packers 69,1 B posts
- 5. Lamar 29,9 B posts
- 6. Jets 53,5 B posts
- 7. #HereWeGo 18,5 B posts
- 8. #GoPackGo 9.052 posts
- 9. Lions 88,4 B posts
- 10. Caleb 31,3 B posts
- 11. Justin Tucker 19,3 B posts
- 12. Taysom Hill 9.335 posts
- 13. Bills 96,5 B posts
- 14. Falcons 9.993 posts
- 15. Paige 9.238 posts
- 16. WWIII 35,2 B posts
- 17. Browns 29,7 B posts
- 18. Russ 14,4 B posts
- 19. Bo Nix 3.167 posts
- 20. Worthy 45,6 B posts
Who to follow
Something went wrong.
Something went wrong.