Search is not available for this dataset
V2 (Flash, frozen) dict | V3 (Llama, new) dict | v2_v3_continuity_rho float64 | v2_v3_common_n int64 |
|---|---|---|---|
{
"dissent_rho": 0.5879815305642176,
"dissent_n": 34,
"hawk_mean": 4.415518768715395,
"dove_mean": -0.7080277957837685,
"n": 4654,
"dict_rho": 0.43823375920136004
} | {
"dissent_rho": 0.6256987381410014,
"dissent_n": 34,
"hawk_mean": 2.838539398012103,
"dove_mean": -1.3701242743673498,
"n": 3937,
"dict_rho": 0.4608340361855826
} | 0.82409 | 3,937 |
Fedlock Archive (private backup)
Off-machine backup of the irreplaceable Fedlock data, created 2026-07-06.
Source of truth lives at C:\Users\Joe\Desktop\fedlock_project (GitHub: jnathan9/fedlock).
Contents
corpus.zip— full speech corpus (19K+ speeches, 1.2GB unzipped)fomc_transcripts.zip— parsed FOMC transcript datatournaments/— all four pairwise-comparison archives (~245K LLM judgments):hawkishness_tournament/— V1, Claude Sonnet 4, mega corpus, 69K comparisonshawkishness_tournament_v2/— V2, DeepSeek V3, FOMC policy corpus, 71Kflash_tournament/— Flash V1, Gemini 2.0 Flash, mega corpus, 60Khawkishness_tournament_flash_v2/— Flash V2 (live dashboard), Gemini 2.0 Flash, 105K comparisons (last ~70 judged by Gemini 2.5 Flash after the 2.0 retirement — see repo notes)
metadata/— corpus indexes, FOMC roster, FRED snapshots, policy classifications, direct-scoring runs, speech registry, duplicate-pair list
The comparison logs are the training data for any future distilled judge model.
- Downloads last month
- 50