Skip to content

Add repeat-level variability statistics to reward profiling - #2474

Open
mpatel31415 wants to merge 33 commits into
mainfrom
mpatelka/add-variability-statistics
Open

Add repeat-level variability statistics to reward profiling#2474
mpatel31415 wants to merge 33 commits into
mainfrom
mpatelka/add-variability-statistics

Conversation

@mpatel31415

@mpatel31415 mpatel31415 commented Aug 11, 2026

Copy link
Copy Markdown
Contributor

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 an
agent 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_metrics as mean_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.

Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
@copy-pr-bot

copy-pr-bot Bot commented Aug 11, 2026

Copy link
Copy Markdown

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>
@github-actions

Copy link
Copy Markdown
Contributor

@mpatel31415 mpatel31415 changed the title feat(eval): initial changes Add repeat-level variability statistics to reward profiling Aug 18, 2026
@mpatel31415
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>
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>
@mpatel31415
mpatel31415 force-pushed the mpatelka/add-variability-statistics branch from 78b9cae to 38bb207 Compare August 25, 2026 07:04
@mpatel31415

Copy link
Copy Markdown
Contributor Author

Non-blocking scope note: #2528 requests three aggregation levels. This PR covers the main repeat-level mechanics, but a few requested outputs remain absent:

  • within-repeat output has sample_count and observed-task missing_count, but no separate valid/failed counts;
  • across-repeat output does not emit the calculated between-repeat standard deviation, min, or max;
  • existing per-task output covers repeat count, mean, median, and standard deviation, but the requested per-task CI is explicitly left out.

I do not think this PR needs to grow to include all of these, but could we state explicitly that it implements a subset of #2528 and leave the remaining pieces for follow-ups? That would prevent the issue from appearing fully satisfied when this merges.

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>
Signed-off-by: Martyna Patelka <mpatelka@nvidia.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

sla:review-overdue Review response is over the one-business-day SLA

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants