Apply explicitly selected dataset corrections

README.md

@@ -240,6 +240,17 @@ configs:
   - split: test
     path:
     - data/test/KVRET.parquet
+- config_name: LMSYS-Chat-1M
+  data_files:
+  - split: train
+    path:
+    - data/train/LMSYS-Chat-1M.parquet
+  - split: validation
+    path:
+    - data/validation/LMSYS-Chat-1M.parquet
+  - split: test
+    path:
+    - data/test/LMSYS-Chat-1M.parquet
 - config_name: MSC
   data_files:
   - split: train
@@ -587,7 +598,7 @@ Roles are already oriented for user-simulator training:
 <!-- available-datasets:start -->
 ## Available Datasets
 
-**Dataset count: 49**
+**Dataset count: 50**
 
 Select an individual dataset by its name below, or use `default` for the combined collection.
 
@@ -606,7 +617,7 @@ Datasets are grouped by behavioral category:
     "llm-facilitation-corpus", "spolin-corpus"
   ],
   "General Conversation & Social Interaction": [
-    "Humanual-Chat", "MSC", "Topical-Chat", "chitchat-dataset"
+    "Humanual-Chat", "LMSYS-Chat-1M", "MSC", "Topical-Chat", "chitchat-dataset"
   ],
   "Information Seeking & Clarification": [
     "FITS", "news-interview", "wizard_of_internet", "wizard_of_wikipedia"

SOURCES.md

@@ -262,3 +262,41 @@ independently established. Links, images, videos and attachments may refer to
 material absent from the released text. Native splits are not claimed to be
 verified leakage-free. Automated source comparison and sampled review validate
 the conversion within these limits.
+
+## LMSYS-Chat-1M (OdysSim subset)
+
+Credit: Lianmin Zheng, Wei-Lin Chiang, Ying Sheng et al.; LMSYS-Chat-1M (2023).
+[Original dataset](https://huggingface.co/datasets/lmsys/lmsys-chat-1m)
+and [paper](https://arxiv.org/abs/2309.11998).
+Prepared source: [OdysSim split release](https://huggingface.co/datasets/Xuhui/sft_processed_large_split/tree/ee332b10508c0a6fbaa86d5de729c89d34f68964/lmsys),
+revision `ee332b10508c0a6fbaa86d5de729c89d34f68964`. This configuration contains the processed OdysSim
+subset, not the complete million-conversation LMSYS release.
+
+Each retained conversation has one human-target perspective. Original human
+`user` messages become `assistant`; original AI `assistant` messages become
+`user`. System context, wording, whitespace, ordering, message boundaries,
+consecutive messages and unanswered human endings are unchanged. No persona
+generation, topic/language filtering, dialogue rewriting or missing-turn
+reconstruction was performed. The three approved shorter training prefixes
+were removed while preserving their longer versions and those versions' personas.
+
+Train: **79,926**; validation: **401**; test: **170**; total: **80,497**.
+These are OdysSim-created splits, with `val` renamed to `validation` and no
+retained conversation reassigned. Supplied personas may be inferred task
+summaries and may expose later dialogue. They are preserved as supplied, not
+verified pre-conversation self-descriptions. Profile fingerprints are not
+authenticated human identities; no guarantee of person-disjoint evaluation
+or general benchmark decontamination is made.
+
+The original [LMSYS dataset license agreement](https://huggingface.co/datasets/lmsys/lmsys-chat-1m#lmsys-chat-1m-dataset-license-agreement)
+contains custom use and redistribution conditions. Separate redistribution
+permission was confirmed by the collection owner for this publication.
+OdysSim's collection-level Apache-2.0 metadata is not used to relicense the
+underlying LMSYS conversations. No blanket license is assigned here.
+
+Complete normalized dialogues and substantive full-prefix overlaps were checked
+against the existing published collection. This does not establish the absence
+of interior shared passages, semantic duplicates or external benchmark items.
+Anonymization placeholders and original AI errors remain in context. Every
+exported message was compared with source; sampled semantic review cannot
+establish the quality or human authorship of every individual record.

data/test/LMSYS-Chat-1M.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94f1d91a8f685bfb25d806f5caf1b6638b86abc28d71dbe0195d1e75e2b4e48b
+size 368180

data/train/LMSYS-Chat-1M.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ead12a9c26eacfd57b8f6b509f11611c348bad4980859a4e64cef3031588bbd
+size 160444712

data/validation/LMSYS-Chat-1M.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a26bd2889be55b6a1b3ecf3b655ef07c5571d47c2f5f037fe50c365347b809e
+size 827893