Spaces:
Running
Running
Commit History
2 days ago
-
fix: four claims the article made that the paper does not
Step 2 described the merge as a bottom-up recursion over equivalent targets and its Note called the relation Myhill-Nerode equivalence, the quotient the unique minimal DFA, and claimed no smaller automaton could reproduce the behavior. The paper merges on the incoming activity, last(x) = last(y), giving at most |A|+1 classes, and appendix.tex:64 states outright that the result is not the minimal DFA of the prefix language, which would be acyclic. The article's own loops and held-out replay contradicted its Note: a Myhill-Nerode quotient of the observed prefixes accepts exactly those prefixes and generalizes to nothing. 59,510 is RPNI's state count from the main table, not the trie's, and the paper gives no trie count for SWE-agent. It appeared twice as the prefix tree size, so both sentences now describe the trie without a number. The 2,380x that came with it is the ratio against RPNI, which the repo forbids citing without the Alergia framing; Alergia needs 35 states against our 25 on…5bc9fb3
12 days ago
-
build: serve the corrected article at the Space root
2d9763d -
prose: split two run-ons, drop the We-find openers, cut a restated clause and give Related Work its third paragraph
7f26f64 -
facts: subsampled sources, the alphabet bound, Alergia and HMM against our count, RPNI's 382 states, four AWM gaps over 15 and the monitor's unspent compute
ce42ee1
14 days ago
-
links: point the paper button at arXiv rather than the workshop OpenReview
The page is now a NeurIPS paper with its own arXiv record, 2608.23670. The variable keeps its name; renaming it is a separate change.77adcae
25 days ago
-
Serve as a static Space
Verified signature664c6e5 -
drop static template stylesheet
Verified signature917df73 -
Serve the built article at the Space root
Verified signature118db0a -
Interactive research article
Verified signature04839fd -
initial commit
Verified signatureda2e5bb