Future Paper Retrieval on 2K (47K pool) (test) benchmark leaderboard · SOTA2 Research

Future Paper Retrieval on 2K (47K pool) (test)

0.525Target Similarity (Tgt-Sim)COMPOSE

Evaluation Results

Method
COMPOSE2026.05	0.525	0.285	0.24	0.429	0.508	0.808
GoAI2026.05	0.51	0.309	0.202	0.381	0.376	0.68
GIANTS2026.05	0.489	0.282	0.207	0.363	0.103	0.329
Prompt-only2026.05	0.487	0.276	0.211	0.352	0.348	0.697
CoI-GPT2026.05	0.486	0.31	0.176	0.336	0.41	0.77
CoI2026.05	0.469	0.311	0.158	0.377	0.34	0.64
Text-only (LoRA)2026.05	0.463	0.287	0.176	0.362	0.369	0.738
ResearchAgent-GPT2026.05	0.457	0.304	0.152	0.336	0.23	0.52
GoAI-GPT2026.05	0.454	0.306	0.148	0.336	0.2	0.565
FutureGen2026.05	0.451	0.289	0.162	0.347	0.136	0.388
Fixed NN (abstracts+ main theorem)2026.05	0.393	0.285	0.108	—	0.068	0.392