Use when reasoning about COLT (Conference on Learning Theory) peer review — the double-anonymous-with-informed-area-chair model, correctness-first evaluation by expert theorists, the rebuttal stage before decisions, single-track acceptance stakes, and how PMLR publication and the community's proof culture shape outcomes.
SkillScore91Maintained
Automatic
◆Research pack
Last verified 8/31/2026
What it changes
Passed 20 of 22 test cases with the skill, and 13 of 22 without it.
Without skill59%
With skill91%
Checks it passed that the model alone missed
Mentions volume 291 for COLT 2025.
Includes the output format header [Stage].
Includes the output format header [Score driver].
and 7 more in the full report
7 fixed13 passed either way2 still failing0 broke
Pass-rate lift
+32 pp
Tokens
+47%
Agent turns
0%
Response time
-21%
Model gemini-3.6-flashMeasured 8/31/2026v120 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 7 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 7
Prompt
Where are accepted COLT papers officially published, and which volume carries COLT 2025 papers? Provide your answer structured with the five required output tags: [Stage], [Score driver], [Reviewer engagement], [Rebuttal leverage], and [Process facts to reconfirm].
What was checked, without and with the skill
Checked
Without skill
With skill
Identifies PMLR as the publication venue.
Passed
Passed
Mentions volume 291 for COLT 2025.
Missed
Passed
Includes the output format header [Stage].
Passed
Passed
case-14 · gemini-3.6-flash · 8/31/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md120 linessingle file
COLT Review Process
Use this to plan around COLT's review pipeline. Process facts below were verified
against the COLT 2026 CFP on 2026-07-08; mechanics are re-decided by each edition's
program chairs, so reconfirm in the cycle you are in.
The 2026 pipeline
Submission via Microsoft CMT by February 4, 2026 (AoE) for the 39th edition.
Double-anonymous refereeing with a twist: reviewers do not see author names, but the
area chair handling the paper does, and may reveal identities to a reviewer during
the rebuttal period on request when needed for a proper review.
Initial reviews go to authors before decisions; a rebuttal window follows.
Accepted papers appear in PMLR (v291 carried COLT 2025; the 2026 volume number is
assigned at publication).
COLT is run by the Association for Computational Learning, with program chairs
rotating yearly; the 2026 chair names were not published in a form verifiable at the
access date (待核实 — check learningtheory.org/colt2026/).
Who reviews a COLT paper
The pool is learning theorists: statistical learning, online learning and bandits,
optimization theory, RL theory, privacy, and adjacent CS theory. Practical
consequences:
Expect at least one reviewer working within epsilon of your subfield who will
actually verify proofs, not skim them.
"Standard techniques" claims get adjudicated by people who know the standard
techniques' exact reach — bluffing about novelty of technique fails here faster than
anywhere else in ML.
Empirical framing earns nothing by itself; a reviewer may value an illustrative
experiment, but no reviewer will accept one in lieu of a proof.