StochStack

prototype 02

Site Feasibility with Human Feedback

Team policy learns from multi-user feedback. Repeated single-user dislikes are capped; weight shifts require consensus.

Base Filter Context

Team Feedback Console

Reason tags

Effective Weights

Experience24.0%
Startup speed18.0%
PI capacity14.0%
Competitor coverage14.0%
CTGov relevance16.0%
Execution stability10.0%
Sponsor fit4.0%

Before (Team Baseline)

#1 Local Institution

Seoul, South Korea

score: 66.2

#2 Local Institution

Hangzhou, China

score: 62.8

#3 Samsung Medical Center

Seoul, South Korea

score: 52

#4 Local Institution

Beijing, China

score: 49.5

#5 Local Institution

Guangzhou, China

score: 49.5

#6 Local Institution

Shanghai, China

score: 49.5

#7 National Taiwan University Hospital

Taipei, Taiwan

score: 38.1

#8 Memorial Sloan Kettering Cancer Center

New York, United States

score: 34.1

After (Team Learned Policy)

#1 Local Institution

Seoul, South Korea

score: 66.2

#2 Local Institution

Hangzhou, China

score: 62.8

#3 Samsung Medical Center

Seoul, South Korea

score: 52

#4 Local Institution

Beijing, China

score: 49.5

#5 Local Institution

Guangzhou, China

score: 49.5

#6 Local Institution

Shanghai, China

score: 49.5

#7 National Taiwan University Hospital

Taipei, Taiwan

score: 38.1

#8 Memorial Sloan Kettering Cancer Center

New York, United States

score: 34.1

Rank Shift

#1 Local Institutionafter #1 (0)
#2 Local Institutionafter #2 (0)
#3 Samsung Medical Centerafter #3 (0)
#4 Local Institutionafter #4 (0)
#5 Local Institutionafter #5 (0)
#6 Local Institutionafter #6 (0)
#7 National Taiwan University Hospitalafter #7 (0)
#8 Memorial Sloan Kettering Cancer Centerafter #8 (0)
#9 Local Institutionafter #9 (0)
#10 Local Institutionafter #10 (0)

Consensus Guardrails

Feedback Log (Audit)

No feedback yet.

    update log

    Prototype Change Log

    1. 2026-03-01 · v0.3.0

      Team Policy Persistence and Consensus Aggregation

      • - Added persistent team profiles and server-side feedback event store.
      • - Added multi-user consensus aggregation for effective weight updates.
      • - Added anti-abuse controls: single-user daily cap, repeated action decay, and max per-reason shift.

    2. 2026-03-01 · v0.2.0

      Human Feedback Reinforcement Layer Added

      • - Added before-vs-after ranking comparison.
      • - Added explicit feedback actions (like/dislike) with reason tags.
      • - Introduced adaptive weight updates and audit feedback log.