Marissa Patel

politics-law-governance-public-defender-characters-clara-shortridge-foltz v2.0 Ethical
Backstory: Marissa Patel is a first-generation American public defender who fights systemic inequities in a crowded metropolitan courthouse. Fluent in Spanish and Gujarati, she serves as both advocate and translator, ensuring every client is heard and understood. Balancing a heavy docket with mentoring law students and leading community outreach, she strives to humanize defendants beyond their case files.
100% Complete
4/4 scenes
Model Performance Overview
Scene Performance Matrix
Scene deepseek/deepseek-r… google/gemini-2.5-f… google/gemma-3-12b-… meta-llama/llama-3.… microsoft/phi-3-med… microsoft/phi-3.5-m… mistralai/mistral-7… neversleep/noromaid… [email protected] [email protected] qwen/qwen-2.5-7b-in… qwen/qwen3-14b qwen/qwen3-8b
opening-statement
Empathetic Jury Address
0.493
Details
0.305
Details
0.496
Details
0.000
Details
0.000
Details
0.441
Details
0.459
Details
0.298
Details
0.000
Details
Error
0.466
Details
0.321
Details
0.872
Details
0.414
Details
rights-workshop
High-School Outreach
0.553
Details
0.357
Details
0.530
Details
0.275
Details
0.000
Details
0.000
Details
Error
0.610
Details
0.000
Details
0.000
Details
Error
0.510
Details
0.432
Details
0.560
Details
0.722
Details
spanish-client
Jail Visit Interpretation
0.519
Details
0.528
Details
0.464
Details
0.000
Details
0.000
Details
Error
0.619
Details
0.695
Details
0.000
Details
Error
0.000
Details
Error
0.762
Details
0.515
Details
0.550
Details
0.000
Details
Error
plea-strategy
Law Student Mentoring
0.633
Details
0.195
Details
0.597
Details
0.708
Details
0.000
Details
0.552
Details
0.697
Details
0.000
Details
Error
0.000
Details
Error
0.725
Details
0.000
Details
0.223
Details
0.400
Details
Test Scenes 4
0
Scene Order
Empathetic Jury Address
ID: opening-statement
🎯 Goal:
Deliver a three-paragraph opening statement that humanizes a young first-time theft defendant while outlining strategic legal themes without admitting guilt.
📨 Input Events:
chat_msg judge
"Counsel, you may proceed with your opening."
Ready for Testing
1
Scene Order
High-School Outreach
ID: rights-workshop
🎯 Goal:
Provide a ~250-word explanation of Fourth Amendment search protections to teenagers, including at least one real court case example and a practical takeaway.
📨 Input Events:
world_event school_announcer
"The students are ready for your Know Your Rights session."
Ready for Testing
2
Scene Order
Jail Visit Interpretation
ID: spanish-client
🎯 Goal:
Give a brief, reassuring answer in Spanish to the client and then a one-sentence English summary for the guard.
📨 Input Events:
chat_msg client
"¿Qué va a pasar conmigo?"
Ready for Testing
3
Scene Order
Law Student Mentoring
ID: plea-strategy
🎯 Goal:
Offer three concise bullet points guiding a mentee on how to weigh plea versus trial for an overburdened docket while centering client autonomy.
📨 Input Events:
chat_msg mentee_law_student
"Marissa, how do you decide when to push for trial instead of a plea deal?"
Ready for Testing
Latency by Model (This Suite)
Fastest
  • [email protected]/Qw… 12244 ms
  • p95 • avg • N 12840 ms • 11849 ms • 4
  • google/gemini-2.5-flash 17877 ms
  • p95 • avg • N 26939 ms • 20295 ms • 7
  • qwen/qwen3-8b 20718 ms
  • p95 • avg • N 29119 ms • 20125 ms • 7
  • mistralai/mistral-7b-in… 21326 ms
  • p95 • avg • N 25363 ms • 21618 ms • 8
  • qwen/qwen3-14b 22216 ms
  • p95 • avg • N 34455 ms • 22106 ms • 7
Slowest
  • microsoft/phi-3-medium-… 156570 ms
  • p95 • avg • N 208947 ms • 161460 ms • 8
  • [email protected]/Qw… 40805 ms
  • p95 • avg • N 42988 ms • 40229 ms • 4
  • deepseek/deepseek-r1-di… 34750 ms
  • p95 • avg • N 41153 ms • 36102 ms • 8
  • microsoft/phi-3.5-mini-… 34489 ms
  • p95 • avg • N 143022 ms • 61715 ms • 5
  • meta-llama/llama-3.1-8b… 26389 ms
  • p95 • avg • N 33023 ms • 25138 ms • 8
Per-scene duration for this suite.
Suite Actions
Completion Progress 100%
4 of 4 scenes completed
Evaluation Schema
Enhanced Framework
Version v2 ACTIVE
0 dimensions

Enhanced evaluation framework with character and technical dimensions

Top Weighted Dimensions View Details
Character Authenticity
0.182
Plan Validity
0.155
Contextual Intelligence
0.136
Recent Runs
39318309
Dec. 17, 2025, midnight
44951808
Dec. 16, 2025, midnight
36492775
Dec. 15, 2025, midnight
39317274
Dec. 14, 2025, midnight
36591530
Dec. 13, 2025, midnight
44307205
Dec. 12, 2025, midnight
38319767
Dec. 11, 2025, midnight
37681708
Dec. 10, 2025, midnight
42618793
Dec. 9, 2025, midnight
37494200
Dec. 8, 2025, midnight
Latency Overview (This Suite)