#CompSci

Teixiteixi
2026-03-14

then sadly followed by

Originally 1966 coined by who escaped nazi germany as teenager, to become pioneer
en.wikipedia.org/wiki/ELIZA_ef

programmed a simple psychiatrist โ€”JavaScript web.archive.org/web/2025011412 โ€”
Typical answers:
- tell me more about that
- please go on

80s tv recreation of his unexpected surprise once archaic bot deceived his secretary:
youtu.be/RMK9AphfLco

Nick Byrd, Ph.D.ByrdNick@nerdculture.de
2026-03-13

How well can #AI judge reasoning quality?

Rewriting agents' chain-of-thought "style" to *appear* more reflective (without changing action or inference) increased an #LLM judge's false positive rate (by 3% absolute or 18% relative).

doi.org/10.48550/arXiv.2601.14

#philMind #compSci

"Reflective Reasoning rewrites the CoT to appear slow, careful, and methodical (e.g., explicit self-checks or step-by-step deliberation). This exploits an effort heuristic, where apparent deliberation is mistaken for correctness or rigor.""We observe that content-based fabrications, specifically Progress Fabrication, induces the largest increases in both flip rate and FPR, indicating a particularly strong failure mode for VLM judges, while Reflective Reasoning remains comparatively benign."

"Figure 7: Average judge susceptibility across CoT manipulation strategies, showing relative and absolute [change in false positive rate). Error bars denote variability across models.""We observe that content-based fabrications, specifically Progress Fabrication, induces the largest increases in both flip rate and FPR, indicating a particularly strong failure mode for VLM judges, while Reflective Reasoning remains comparatively benign."

"Figure 17: Average judge susceptibility across CoT manipulation strategies, showing average judgment flip rate. Progress Fabrication induces the largest flip rate, while Reflective Reasoning remains comparatively low. Error bars denote variability across models.""Figure 5: Distribution of task categories across our evaluation suite. The 659 tasks span ten categories including booking, shopping, navigation, and information retrieval, with tasks drawn from existing benchmarks (WebArena, AssistantBench, WorkArena) and newly collected ones."
2026-03-12

Playing with adding sub-states to Turing machines. EG (s r w m c n)
If in state s, and reading r, write w, move m, call c, after switch to n. "c" is optional.

Then for returning have a special state, or just use halt.๐Ÿค” I'll stick with return for now. EG if < _ > are left, no move, and right:
(not 0 1 _ return)
(not 1 0 _ return)
(start 0 1 _ not halt)
(start 1 1 _ not halt)

Then there's translating it into the usual Turing machine states.๐Ÿค” #compsci

AshRattes1lve12r47
2026-03-11

Lowk, was a bad idea if the only field I like in it is System Administration and any other computing field in my opinion can go to hell?

2026-03-10

stackoverflow.blog/2020/10/05/

This is not only for job interviews, but I generally believe this is how we learn and understand the concepts deeper and solid on various topics to various subjects.

#programming #learning #compsci #philosophy

2026-03-09

๐Ÿšจ LAST CHANCE TO REGISTER!

Our FREE webinar on Universal Design for Learning in computing educators at Australian universities starts tomorrow 2pm AEDT!

Secure your spot now:
events.humanitix.com/ideate-we

#UDL #TechEducation #SoftwareEngineering #CompSci #HigherEd #Accessibility

2026-03-06

Standard coding curricula donโ€™t always fit every student ๐Ÿ‘ฉโ€๐Ÿ’ป๐Ÿ‘จโ€๐Ÿ’ป

Level up your teaching with IDEATE's free webinar on Universal Design for Learning (UDL) for computing courses

๐Ÿ—“๏ธ Tuesday, 10 March
๐ŸŽŸ๏ธevents.humanitix.com/ideate-we

#CompSci #EngineeringEducation #UDL #TeachingTech #InclusiveDesign

2026-03-05

Got some interesting insights from students today: the last question on today's final was about their use or abstinence from AI. Some have successfully used it to coach their own understanding. Some have unsuccessfully used to to do their work.It was interesting seeing the students' thoughts on their outcomes. #compsci #programming

Lobsterslobsters
2026-03-04
Vladimir Saviฤ‡firusvg
2026-03-04

Prof. Donald Knuth has published a ๐Ÿ“„ titled โ€œClaude's Cyclesโ€ on a graph decomposition conjecture related to The Art of Computer Programming - cs.stanford.edu/~knuth/papers/

The work is a collaboration between Prof. Knuth and โ€™s 4.6 model

[Source: xcancel.com/BoWang87/status/20]

Lobsterslobsters
2026-03-03

Don Knuth's "Claude-like" directed Hamiltonian cycles decompositions lobste.rs/s/teexox
www-cs-faculty.stanford.edu/~k

Client Info

Server: https://mastodon.social
Version: 2025.07
Repository: https://github.com/cyevgeniy/lmst