Skip to content
Danphe Council

About

Why Danphe Council exists

Single AI models write fluent answers that are sometimes wrong in ways that are hard to spot. We built Danphe so that people who are accountable for an answer can see which parts hold up.

The idea

The danphe, the Himalayan monal, is Nepal's national bird, and the mark of Danphe Intelligence Lab, the research company in Kathmandu that builds Danphe Council.

Danphe Council asks several leading models from different companies the same question, has them check each other's statements against your documents, and keeps a tamper-evident record of what happened. Where they disagree, you see it.

We test this the way it should be tested: on questions the system has never seen, with the rules written down first. The results are on the accuracy page.

Built by
Danphe Intelligence Lab
Founded
2024, Kathmandu, Nepal
Research focus
Calibration, uncertainty, abstention and LLM evaluation
Legal entities
Danphe Intelligence Lab Pvt. Ltd. (Nepal) and Danphe Intelligence Lab LLC (USA)

danphelab.com

Who's behind this

Danphe Intelligence Lab

An AI research company whose work on calibration and uncertainty is the reason Danphe Council labels, quotes and measures instead of just answering.

Built by Danphe Intelligence Lab

Danphe Intelligence Lab is an AI research company founded in Nepal in 2024. Its research is on making machine-learning systems say how sure they are, and be right about it. That is the question behind every label in Danphe Council.

Besides Danphe Council, the lab builds a biosignal analysis API with calibrated uncertainty and an open-source ML calibration and uncertainty toolkit, and takes on fixed-scope research work such as calibration audits.

  • Kathmandu, Nepal
  • Danphe Intelligence Lab Pvt. Ltd. (Nepal) and Danphe Intelligence Lab LLC (USA)
danphelab.comGitHub[email protected]

Research behind the labels

  • Calibration

    Measuring whether a model's confidence matches how often it is right (ECE, temperature scaling).

  • Uncertainty and abstention

    Knowing when a system should not answer, and saying so, instead of guessing.

  • LLM evaluation

    Preregistered tests on unseen data, with the rules written down before the run.

What we hold ourselves to

  • Show the work

    An answer is only as useful as your ability to check it. Every statement comes with its source and its verdict.

  • Say what we measured

    We publish our accuracy results with sample sizes, dates and method, including where we are not better.

  • No single vendor

    Models from different companies check each other, and none checks its own family's statements.

Book a demo

Design partner programme

3 places for legal, audit or compliance teams

3 months of Pro Max free, in exchange for feedback while we shape the product with you.

Apply as a design partner