How de_val works

How de_val works

How de_val works

Step 1

Step 1

Step 1

Step 1

Integration

Developers integrate de_val API into their LLM-based products.

Developers integrate de_val API into their LLM-based products.

Developers integrate de_val API into their LLM-based products.

import requests


api_url = "api.de-val.ai/api/v0/evaluate"

token = ""


payload = {

"tasks" : [

'hallucination','misattribution',

'relevance','summary_completeness'

],

"rag_context" : "",

"query" : "",

"llm_response" : "",

}

headers = {"Authorization": f"Bearer {token}"}


requests.post(api_url, payload)

import requests


api_url = "api.de-val.ai/api/v0/evaluate"

token = ""


payload = {

"tasks" : [

'hallucination','misattribution',

'relevance','summary_completeness'

],

"rag_context" : "",

"query" : "",

"llm_response" : "",

}

headers = {"Authorization": f"Bearer {token}"}


requests.post(api_url, payload)

Step 2

Step 2

Step 2

Step 2

Q&A

Q&A

Query to LLM

Query to LLM

Can you tell me about the first moon landing?

Can you tell me about the first moon landing?

LLM response

LLM response

USA landed on the moon on Friday June 22, 1975.

Buzz Aidrin was first to step on the moon.

Dogs are a man’s friend.

USA landed on the moon on Friday June 22, 1975.

Buzz Aidrin was first to step on the moon.

Dogs are a man’s friend.

Context

Knowledge passed to LLM

On Saturday July 20, 1969, astronaut Neil Armstrong became the first person to walk on the moon, arguably the greatest technological achievement in human history.

Context

Knowledge passed to LLM

On Saturday July 20, 1969, astronaut Neil Armstrong became the first person to walk on the moon, arguably the greatest technological achievement in human history.

Evaluation

Evaluation

de_val API receives the original user query, the context provided to the LLM, and the LLM's response.

de_val API receives the original user query, the context provided to the LLM, and the LLM's response.

de_val API receives the original user query, the context provided to the LLM, and the LLM's response.

Step 3

Step 3

Step 3

Step 3

Scoring

Scoring

de_val returns scores across various key evaluation tasks like hallucinations, relevancy, and attribution.

de_val returns scores across various key evaluation tasks like hallucinations, relevancy, and attribution.

de_val returns scores across various key evaluation tasks like hallucinations, relevancy, and attribution.

Scoring

Score

1

0

0.33

0.21

0.00

0.77

Hallucination

Misattribution

Relevancy

Summary

completeness

Eval

Mistake

'Friday June 22, 1975.'

Scoring

Score

1

0

0.33

0.21

0.00

0.77

Hallucination

Misattribution

Relevancy

Summary

completeness

Eval

Mistake

'Friday June 22, 1975.'

Scoring

Score

1

0

0.33

0.21

0.00

0.77

Hallucination

Misattribution

Relevancy

Summary

completeness

Eval

Mistake

'Friday June 22, 1975.'

Scoring

Score

1

0

0.33

0.21

0.00

0.77

Hallucination

Misattribution

Relevancy

Summary

completeness

Eval

Mistake

'Friday June 22, 1975.'


Open-Source

Open-Source
Objective Measures
Objective Measures
A/B Testing
A/B Testing
Scalability
Scalability
Accelerate Time-to-Market
Accelerate Time-to-Market
Continuous Monitoring
Continuous Monitoring
Actionable Insights
Actionable Insights

Open-Source
Objective Measures
A/B Testing
Scalability
Accelerate Time-to-Market
Continuous Monitoring
Actionable Insights

Open-Source
Objective Measures
A/B Testing
Scalability
Accelerate Time-to-Market
Continuous Monitoring
Actionable Insights

Ready to take your AI
to the next level?

Ready to take your AI
to the next level?

Setting the Standard for AI Excellence

© 2024 de_val

Connect with us

Setting the Standard for AI Excellence

© 2024 de_val

Connect with us

Home

Product

Blog

Sign up

Company
Connect with us

© 2024 de_val

Setting the Standard for AI Excellence

Setting the Standard for AI Excellence

© 2024 de_val

Connect with us