Runs - Product Search and Recommendation 2025¶

baseline_run¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: baseline_run
Participant: UIUC
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-11
Task: trec2025-product-search
MD5: ad9b81c314756ebfe0df6475cb202c44
Run description: Baseline method using BM25 to complete the task without any query reformulation.

bm25¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: bm25
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: 2294326e235f5e3ab1dcf78ec60fdf35
Run description: bm25

gar_rm3_f120d10w3¶

Run ID: gar_rm3_f120d10w3
Participant: DUTH
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: 4f70af45b103e110b5fd6d1aa444ec88
Run description: BM25+RM3 (fb_terms=120, fb_docs=10, original_query_weight=0.3). Top-1000 per qid.

garamp_bm25_v1¶

Run ID: garamp_bm25_v1
Participant: DUTH
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: cd4d853b9ab59d01709f030a9782e3e4
Run description: BM25 (k1=0.9, b=0.4). Top-1000 per qid

garamp_prf_v1¶

Run ID: garamp_prf_v1
Participant: DUTH
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: 8411b0377ad1ce4e6391d7a07b661168
Run description: Interactive: 4 PRF-based reformulations/query mined from BM25 top-50; BM25 retrieval (k1=0.9, b=0.4). Automatic; no manual curation. Top-1000 unique per qid.

garamp_rm3_f40d5_w05¶

Run ID: garamp_rm3_f40d5_w05
Participant: DUTH
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: fe85fd4f5ede4057b6ecad1849b8fceb
Run description: BM25+RM3 (fb_terms=40, fb_docs=5, original_query_weight=0.5). Top-1000 per qid.

garamp_rm3_v1¶

Run ID: garamp_rm3_v1
Participant: DUTH
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: 0aae6fe5a983ac03faa82865d636fd7a
Run description: BM25+RM3 (fb_terms=80, fb_docs=8, original_query_weight=0.3). Top-1000 per qid.

jbnu-r01¶

Participants | Proceedings | Input | recsys-eval | Appendix

Run ID: jbnu-r01
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-rec
MD5: d40a40c9f46dc0f490f04a7eadae0f90
Run description: 1. First-stage
ColBERT trained on substitute–complement triplets
Classification/ Rerank (Boost)
Qwen3-14B few-shot classification / rerank (if substitute, multiply substitute score by 10; if complement, multiply complement score by 10)

jbnu-r02¶

Participants | Proceedings | Input | recsys-eval | Appendix

Run ID: jbnu-r02
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-rec
MD5: b30e860e0b9d737eef0ca3f9f41fe236
Run description: 1. First-stage
ColBERT non-trained
Classification/ Rerank (Boost)
Qwen3-14B few-shot classification / rerank (if substitute, multiply substitute score by 10; if complement, multiply complement score by 10)

jbnu-r03¶

Participants | Proceedings | Input | recsys-eval | Appendix

Run ID: jbnu-r03
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-rec
MD5: aa1c09d340fa3950464e5b98c79e15dc
Run description: 1. First-stage
ColBERT non-trained
Classification/ Rerank (Boost)
ModernBERT classifier trained on S/C/I labels / rerank (if substitute, multiply substitute score by 10; if complement, multiply complement score by 10)

jbnu-r04¶

Participants | Proceedings | Input | recsys-eval | Appendix

Run ID: jbnu-r04
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-rec
MD5: 136428cd592a738a0cf3a892cf185f05
Run description: 1. First-stage
ColBERT trained on substitute–complement triplets
Classification/ Rerank (Boost)
ModernBERT classifier trained on S/C/I labels / rerank (if substitute, multiply substitute score by 10; if complement, multiply complement score by 10)

jbnu-r05¶

Participants | Proceedings | Input | recsys-eval | Appendix

Run ID: jbnu-r05
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-rec
MD5: eab72af3e9bc5fcdb1e633316b72730e
Run description: 1. First-stage
SPLADE non-trained
Classification/ Rerank (Boost)
ModernBERT classifier trained on S/C/I labels / rerank (if substitute, multiply substitute score by 10; if complement, multiply complement score by 10)

jbnu-s01¶

Run ID: jbnu-s01
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: 26daa7e528b2aaa81f25bd54283bfaf5
Run description: Query Expansion Method
Qwen3-14B (LLM) used to generate Lucene search queries with CoT, ToT, user intent analysis, and reasoning for query expansion

jbnu-s02¶

Run ID: jbnu-s02
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: 5ae2849a9a4c304d7de30911da62eb68
Run description: Query Expansion Method
Qwen3-14B (LLM) used to generate Lucene search queries with ToT-based query expansion

jbnu-s03¶

Run ID: jbnu-s03
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: d1ddb796323824255632a0dd0dc8cd7e
Run description: Query Expansion Method
Qwen3-14B (LLM) used to generate Lucene search queries with synonym-focused query expansion

jbnu-s04¶

Run ID: jbnu-s04
Participant: JBNU
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-14
Task: trec2025-product-search
MD5: dc2663007287481b3cf5d6140db728f9
Run description: Query Expansion Method
Collected 3 related images from Google search, then used Qwen2.5-VL (VLM) for image recognition to expand Lucene search queries

llmbaseline_g27b¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: llmbaseline_g27b
Participant: UIUC
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-11
Task: trec2025-product-search
MD5: cfbb536da81c942fc44d38f659b17616
Run description: A simple gemma 27b based query reformulation. The LLM is prompted to expand the query to improve bm25 performance by anticipating similar product names or keywords to the query.

multiagent_g27b¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: multiagent_g27b
Participant: UIUC
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-12
Task: trec2025-product-search
MD5: 75633ebecf7a614fad3491960357894b
Run description: This run uses a multi-agent approach to first attempt to understand user intent, then produce plausible keywords and products, and finally synthesize prior steps with original query.

reform_iter1¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter1
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: 7249e2a54dfa24c6bac82c5cdaa06dba
Run description: blind reformulation using an llm, agnostic of the catalogue

reform_iter10¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter10
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: f86008ce13f27a79d370c6f295350c17
Run description: diversity focused content rerank of iter 3

reform_iter11¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter11
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: 63e2a3813717fa4b926d4c776b6e3a39
Run description: top 30 relevenace focused diversity rerank of iter 3

reform_iter3¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter3
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: c8903541269492e97a6368c19ba1424b
Run description: better reformulations using llms with product type understanding of the catalogue

reform_iter6_final¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter6_final
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: e0960fb5b1a786081b0bf6c30a0065d9
Run description: agentic approach, iterative improvement using llm judgment

reform_iter8¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter8
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: 6f68b56fc42aca7f4bd36dd358952c07
Run description: diversity focused agentic reformulation

reform_iter9¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_iter9
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: 9630fa3cc7a3e244d7c4c79037fd1da8
Run description: iter 3+ rerank - id based diversity

reform_test2¶

Participants | Input | product_search.trec_eval | trec_eval | Appendix

Run ID: reform_test2
Participant: Trec-PSRT-Orgs
Track: Product Search and Recommendation
Year: 2025
Submission: 2025-09-15
Task: trec2025-product-search
MD5: a61f7134f096c81201caa1b62ff41753
Run description: query level reformulations