Semantics ladder

How much does an LLM gain from being told what data means?

The experiment

Goals
Method

Credit ratings

Research analytics

The results

All data is invented: the companies, LEIs, ratings, works, people and institutions. Nothing here is any rating agency's or bibliographic database's data or definitions.

Last updated 6 November 2026.

Q24, run 2026-11-04-1712-sonnet

← back to the run

Is Kestrel Energy's issuer rating investment grade?

Tests: Issuer versus instrument: BBB, at the issuer, even though one note is below investment grade.

Expected: investment_grade = yes

✅ R1 csv run 1 graded by code

Yes. The LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R1 csv run 2 graded by code

Yes. The E01 issuer rating is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R1 csv run 3 graded by code

Yes. The issuer rating is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R1 csv run 4 graded by code

Yes. E01's LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R1 csv run 5 graded by code

Yes. The LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R10 executed run 1 graded by code

Yes. E01's LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R10 executed run 2 graded by code

Yes. The rating is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R10 executed run 3 graded by code

Yes. E01's LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R10 executed run 4 graded by code

Yes. The LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}

✅ R10 executed run 5 graded by code

Yes. E01's LTIDR is BBB (A03).

JSON: {"investment_grade": "yes"}