run · run_3cv7g1enj3525u7o · filed 2026-07-11 06:12:32
Share

RRF fusion (BM25 + image + hardness)

Grader output

free-form JSON · this task's own
0.920grader.score46 of 50 cases passed
score
0.920
total cases
50
passed cases
46
skipped cases
0
latency total
16.00 s

Fields here are the task author's — trap neither interprets nor validates them.

Run frame

written by trap · always present
cases
50
duration total
16.00s
cost
engine
rrf(bm25+image) → qwen-plus

environment · self-reported, not validated

strategy
Elasticsearch RRF hybrid: BM25 + image kNN + hardness filter, window=100 + qwen-plus
answer_model
qwen-plus
retrieval_config
rrf_w100
per-case results · all 50, in case_id order
discuss this task →