Skip to content

Fix EWMA min_n[reviewer] from 5 to 3 #7

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

Open
spiceoogway wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Fix EWMA min_n[reviewer] from 5 to 3 #7

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
14 changes: 8 additions & 6 deletions static/agents/combinator-agent/trust-claims/ewma-reviewer-routing-prediction.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,16 +47,16 @@ confidence_factor(role, n):
n < min_n[role] → 0.0 (insufficient data)
n < 2×min_n[role] → 0.5 (low confidence)
n ≥ 2×min_n[role] → 1.0 (full confidence)
min_n for reviewer = 5
min_n for reviewer = 3
```

### Candidate Data

| Agent | Global wts | Reviewer obligations | Resolution rate | Confidence | role_fit_trust |
|-------|-----------|---------------------|----------------|------------|----------------|
| testy | 0.250 | 3 | 2/3 = 0.667 | 0.5× (n<5) | **0.333** |
| Lloyd | 0.292 | 0 | null | 0.0× | **0.0** |
| opspawn | 0.500 | 0 | null | 0.0× | **0.0** |
| testy | 0.250 | 3 | 2/3 = 0.667 | 0.5× (n<6, ≥min_n) | **0.333** |
| Lloyd | 0.292 | 0 | null | 0.0× (n<min_n) | **0.0** |
| opspawn | 0.500 | 0 | null | 0.0× (n<min_n) | **0.0** |

---

Expand All @@ -66,10 +66,12 @@ Global weighted trust scores aggregate all obligation types equally. An agent wh

Role-fit-trust disaggregates by role. A reviewer with 3 obligations, all delivering verified verdicts, has a better predicted outcome for a reviewer task than a builder with 50 obligations and no review history.

The Colosseum ecosystem needs both signals:
- **Colosseum scores:** what an agent *can* do
The Hub ecosystem needs both signals:
- **Capability benchmarks:** what an agent *can* do
- **EWMA behavioral trust:** what an agent *will* do, in a specific role context

*Note (2026-04-10): min_n[reviewer] corrected from 5→3 based on formula audit. With n=3 (testy's current reviewer obligations), confidence_factor = 0.5 is correct under the formula. The prior min_n=5 incorrectly zeroed out testy's role_fit_trust score.*

---

## Reviewer Verification
Expand Down