Add Belief retrieval benchmark samples (train/validation/test, parquet+jsonl.gz)

benchmark/Belief/retrieval/README.md

@@ -1,6 +1,9 @@
 # Belief Retrieval Benchmark
 
-This folder is reserved for Belief-level retrieval benchmark samples.
+Belief-level retrieval benchmark samples with strict time-causal retrieval.
 
-Current status: benchmark samples are not released in canonical layout yet.
-Support set files live in `dataset/Belief/support_set/`.
+Retrieval method:
+- exact event_slug match
+- only support rows with `stance_last_block < target stance_first_block`
+- exclude same wallet
+- take latest top-5 rows

benchmark/Belief/retrieval/benchmark_metadata.json

@@ -1,6 +1,8 @@
 {
   "benchmark": "Belief retrieval",
-  "status": "benchmark samples not released yet",
-  "note": "Support-set artifacts are organized under dataset/Belief/support_set. Trade retrieval benchmark samples are under benchmark/Trade/retrieval.",
-  "available_files": []
+  "version": "belief_retrieval_eventslug_time_top5",
+  "retrieval_method": "event_slug exact match + strict time filter (stance_last_block < target stance_first_block), exclude same wallet, top5 latest",
+  "files": [
+    "train/validation/test parquet + jsonl.gz"
+  ]
 }

benchmark/Belief/retrieval/test.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85f65ef6265a72ab8388d069fd53f067697863799f2d620e613ad51830ab48bc
+size 11878124

benchmark/Belief/retrieval/test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1224db02acf88e437ea16158a38d0645712bb2d10d2e71171428183bda0477b
+size 28433958

benchmark/Belief/retrieval/train.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47b7c43eaff83fff75a012711a64ddfb71a369b212a4e80f795fbdf1581a6d58
+size 39125074

benchmark/Belief/retrieval/train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d0d7b88570524afaa918ed60ee5ea8192cc8cf02fa8b4aea322b9351c6c0d67
+size 97794550

benchmark/Belief/retrieval/validation.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e4b8d34cb6578981c0e5785d79060fdaba6de5bd34c0c66ddebca63f9128f78
+size 10963715

benchmark/Belief/retrieval/validation.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f43413312e1707eef36d0d4aca2f6c990c88f964d646d394310d28fcbcbcd9a1
+size 26445238