Twitxy Saved

Ian Arawjo

@IanArawjo

Asst Prof @UMontreal, PhD @CornellInfoSci. Creator of @chainforge_ai and evalstats. HCI, programming, and LLM evaluation. Prev: Postdoc @HarvardHci
Main Memory linktr.ee/ianarawjo Joined May 2011
1.1K Following 2.7K Followers 10K Posts

Latest posts 5

Pinned post

Ian Arawjo @IanArawjo
We are hiring! I am looking to hire a PhD student to work on topics around AI evaluation from an HCI and technical standpoint. If you are interested, please apply through the Mila portal and put my name down as a potential advisor:
Mila - Institut québécois d'IA Verified account @Mila_Quebec Oct 15
Mila's annual supervision request process is now open to receive MSc and PhD applications for Fall 2027 admission! For more information, visit mila.quebec/en/prospective…
View quoted post on X
0 replies 14 reposts 57 likes 3.9K views View on X
Ian Arawjo @IanArawjo
The reception to this has been amazing so far! But we need your submissions. Submit your reviews to HCI venues, past or present—accept or reject or an ADR decision. Let’s build up data on what HCI reviews look like and how quality may be changing!
Ian Arawjo @IanArawjo 5d
Introducing OpenCHIReviews, a crowdsourced database of reviews for ACM HCI conferences, chiefly CHI. Submit anonymized reviews and paper decisions, and see what other HCI papers' reviews look like, in the spirit of open science:
View quoted post on X
4 replies 3 reposts 12 likes 844 views View on X
Ian Arawjo @IanArawjo
Introducing OpenCHIReviews, a crowdsourced database of reviews for ACM HCI conferences, chiefly CHI. Submit anonymized reviews and paper decisions, and see what other HCI papers' reviews look like, in the spirit of open science:
6 replies 23 reposts 111 likes 9K views View on X
Ian Arawjo @IanArawjo
It’s so great to see community analysis and postmortems of peer review in their communities. If only HCI had a similar ability…
Weirui Ye Verified account @Weirui_Ye 6d
I scraped all 38,101 ICLR 2027 submissions and counted the names before the colon.

LLM has a naming crisis (they should have memory in visit counts of naming).

🏆 The leaderboard:
TRACE: 47 papers
PRISM: 42
PACE: 22
SCOPE: 21
SAGE: 17
MOSAIC: 16
COMPASS: 14

750 names are used by 2+ papers. That's 2,209 papers sharing a name.

Best parts:

• Two papers called TRACE both fix VLM hallucination. Same name, same problem, different authors.

• Four papers called PRISM do time series. Good luck, reviewers.

• "PACE: Discovering Options via Self-Distillation." None of P, A, C, E appear in the title. They picked the name first.

• 47 TRACEs cover quantum computing, materials discovery, seizure localization, carbon emissions, traffic control, and circuit design. TRACE is everywhere.

• The formula: 5 letters, looks like a real word, sounds cool. CAST. PACT. CRAFT. GRAFT. FORGE.

Tip for XXXX 2027: search OpenReview before you name your method 😂

Follow for more data-backed ML nonsense.
View quoted post on X