haiku.rag/evaluations/configs
Yiorgis Gozadinos 73578a1198
Add the pooled four-domain retrieval dataset
`mtrag_federated` partitions one domain by article title, which is
round-robin fusion's friendliest case: no collection is ever off-topic for a
query, so the guaranteed-slot waste that hurts a real deployment is never
exercised. Every fusion conclusion measured on it is therefore provisional.

`mtrag_pooled` pools all four MTRAG domains, so a query belongs to one and
the rest are genuinely off-topic. `collection_of` gains `alpha`, which now
means something: 0 keeps a collection to one domain, 1 ignores the domain
and shards titles uniformly. Domains map onto collections proportionally,
subdividing by title where there are more collections than domains and
grouping where there are fewer.

Passage ids are checked for collisions across domains, since gold is
uri-keyed and a shared id would make it ambiguous.

Claude-Session: https://claude.ai/code/session_01WhudUtZm6qqiuv8Y1sbwSc
2026-08-31 19:08:58 +03:00
..
frames.yaml Point FRAMES at the live reranker and give the judge room to think 2026-08-24 09:03:44 +03:00
hotpotqa.yaml Pin the eval judge to qwen3.8 2026-08-18 14:28:16 +03:00
mtrag_clapnq.yaml Pin the eval judge to qwen3.8 2026-08-18 14:28:16 +03:00
mtrag_federated.yaml Add the federated ClapNQ retrieval dataset 2026-08-31 19:08:58 +03:00
mtrag_pooled.yaml Add the pooled four-domain retrieval dataset 2026-08-31 19:08:58 +03:00
orb_multimodal.yaml Pin the eval judge to qwen3.8 2026-08-18 14:28:16 +03:00
orb_multimodal_nemotron.yaml Pin the eval judge to qwen3.8 2026-08-18 14:28:16 +03:00
orb_text.yaml Pin the eval judge to qwen3.8 2026-08-18 14:28:16 +03:00
t2_finqa.yaml Pin the eval judge sampling and standardise on Qwen3-Reranker 2026-08-06 13:17:58 +03:00