Add repeat-level variability statistics to reward profiling - #2474
Open
mpatel31415 wants to merge 18 commits into
Open
Add repeat-level variability statistics to reward profiling#2474mpatel31415 wants to merge 18 commits into
mpatel31415 wants to merge 18 commits into
Conversation
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
…d-variability-statistics Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Contributor
|
🌿 Preview your docs: https://nvidia-preview-mpatelka-add-variability-statistics.docs.buildwithfern.com/nemo/gym Here are the markdown pages you've updated: |
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
mpatel31415
changed the title
mpatel31415
marked this pull request as ready for review
August 18, 2026 09:51
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
…ability-statistics' into mpatelka/add-variability-statistics Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
github-actions
Bot
added
the
sla:review-overdue
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Issue this PR addresses
Related new bug
Adds a per-repeat statistics to reward profiling / aggregation, so a score can be
reported with a measure of how much it moves between whole runs rather than as
a bare point estimate.
What's new
repeat_level_metrics— one entry per(agent, repeat), computed when anagent has two or more completed repeats. Each entry carries
sample_count,missing_count, and per numeric field:mean,median,std(ddof=1),sem,min/max,p25/p75, and a 95% Student's-t confidence interval.Cross-repeat aggregates — the per-repeat means are summarized across repeats
and merged into
agent_metricsasmean_across_repeats/mean/{field},median_across_repeats/…,se_across_repeats/…, and the 95% CI bounds.Note
Per task across samples aggregates deliberately gets no CI — a per-task distribution over
repeated outcomes can't be assumed normal, and the CLT doesn't apply to raw
outcomes, so their computation / formula is benchmark and metric specific.