How Bonus Points and Deductions Work in Hiring Agent Scoring

Hiring Agent adds up to 20 bonus points for exceptional achievements and subtracts deductions for deficiencies, combining these with category scores to produce a final composite score capped at the maximum possible threshold.

The interviewstreet/hiring-agent repository implements a nuanced evaluation system that grades technical candidates across multiple dimensions. Understanding how bonus points and deductions modify the base category scores is essential for interpreting evaluation results and integrating the scoring API into your hiring pipeline.

Understanding Bonus Points and Deductions in Hiring Agent

Bonus Points Structure

In main/models.py, the BonusPoints class defines how extra credit is awarded. The total field accepts values between 0 and 20, representing points added to the raw category scores. The breakdown field stores a free-form string explaining the specific achievements that earned the bonus, such as exceptional open-source contributions or leadership roles.

Deductions Structure

Also defined in main/models.py, the Deductions class penalizes missing or substandard items. The total field stores a positive integer representing points to subtract, while the reasons field documents the justification—for example, "Missing portfolio link" or "Incomplete technical documentation." These points are subtracted from the sum of category scores and bonus points.

The Scoring Algorithm in score.py

The print_evaluation_results function in main/score.py implements a four-step calculation:

1. Calculate capped category scores for Open-Source, Self-Projects, Production, and Technical Skills.
2. Add bonus points from evaluation.bonus_points.total to the running sum.
3. Subtract deductions using evaluation.deductions.total from the total.
4. Clamp the result to the maximum possible overall score, defined as max_score + 20 (accounting for the 20-point bonus cap).

The final output displays the composite score alongside the bonus and deduction breakdowns.

Code Implementation Examples

Creating Evaluation Data with Custom Scoring

You can construct EvaluationData objects directly for testing or integration purposes:

from models import (
    CategoryScore,
    Scores,
    BonusPoints,
    Deductions,
    EvaluationData,
)

# Define capped category scores

scores = Scores(
    open_source=CategoryScore(score=30, max=35, evidence="Contributed to 3 OSS projects"),
    self_projects=CategoryScore(score=25, max=30, evidence="Built 2 personal apps"),
    production=CategoryScore(score=20, max=25, evidence="2 years at Acme Corp"),
    technical_skills=CategoryScore(score=8, max=10, evidence="Proficient in Python, Go")
)

# Award up to 20 bonus points

bonus = BonusPoints(total=12, breakdown="Extra points for open-source leadership")

# Apply deductions (stored as positive, applied as negative)

deductions = Deductions(total=3, reasons="Missing portfolio link")

# Assemble evaluation

evaluation = EvaluationData(
    scores=scores,
    bonus_points=bonus,
    deductions=deductions,
    key_strengths=["Strong problem-solving", "Team player"],
    areas_for_improvement=["Documentation", "Testing"]
)

Running Evaluations from the Command Line

For standard usage, process résumés through the CLI:

python score.py path/to/candidate_resume.pdf

The script extracts content, invokes the LLM evaluator, and prints formatted results including the bonus and deduction details.

Summary

• Bonus points are defined in main/models.py with a hard cap of 20 points and require a textual breakdown of achievements.
• Deductions subtract from the total score and include documented reasons for the penalty.
• The scoring pipeline in main/score.py applies bonuses additively and deductions subtractively before clamping to the maximum threshold.
• Both components integrate into the EvaluationData structure returned by the LLM evaluator.

Frequently Asked Questions

What is the maximum number of bonus points in Hiring Agent?

The BonusPoints class enforces a maximum of 20 points via validation logic in main/models.py. This limit ensures the bonus system remains a secondary modifier rather than overriding the core category evaluation.

How are deductions represented in the data model?

Deductions store the point value as a positive integer in the total field but are applied as a negative value during the final calculation in print_evaluation_results. The reasons field provides transparency for why points were removed.

Can the final score exceed the standard maximum?

Yes, but only by the bonus cap. The final score clamps to max_score + 20, meaning if a candidate scores perfectly across all categories and receives 20 bonus points, they achieve the theoretical maximum.

Where does the EvaluationData object originate?

The EvaluationData object containing scores, bonuses, and deductions is generated by the LLM evaluator (typically in main/evaluator.py) and consumed by the scoring logic in main/score.py to produce the final composite grade.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}