Skip to content

Methods: fill the abstention rate placeholder from the completed run #291

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
jplfaria wants to merge 1 commit into from
Closed

Methods: fill the abstention rate placeholder from the completed run #291

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 7 additions & 6 deletions Papers/NAR_Update_2026/latex/sections/methods_council_direction.tex
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,6 @@
%% Direction assignment from reaction chemistry (LLM ensemble).
%% \TBD values pending the full-database run, in progress.
%% Run complete over all 45,644 reactions. The one remaining \TBD is the
%% supplementary reference, pending final numbering.
\subsection{Direction assignment from reaction chemistry}\label{sec:methods-council}

Thermodynamic direction assignment is bounded by estimate coverage: a reaction
Expand Down Expand Up @@ -53,11 +54,11 @@ \subsection{Direction assignment from reaction chemistry}\label{sec:methods-coun
so individual assignments can be reviewed rather than accepted in aggregate.
Two properties of the method bear on how the calls should be used. The
ensemble commits to a direction far more readily than the thermodynamic rules
do, abstaining on \TBD[pct]\% of reactions where the rules decline on a
substantial fraction of the database, so it supplies calls where none
currently exist rather than reproducing the ones that do. And its stated
confidence does not distinguish its correct calls from its incorrect ones, so
confidence is not a usable filter for downstream integration. A systematic
do, abstaining on 1.7\% of the reaction set where the existing rules decline
on 48.7\% of it, so it supplies calls where none currently exist rather than
reproducing the ones that do. And its stated confidence does not distinguish
its correct calls from its incorrect ones, so confidence is not a usable
filter for downstream integration. A systematic
evaluation against measured equilibrium constants, and against the
model-level consequences of adopting these calls, is deferred to follow-up
work.