AdarshSingh7647/Std-SFT-Passage-14B

TEXT GENERATIONPricing:Input $0.48 / Output $0.96Concurrent Unit Cost:1Model Size:14BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 26, 2026Architecture:Transformer Featherless Exclusive Cold

AdarshSingh7647/Std-SFT-Passage-14B is a 14 billion parameter Qwen3-based passage reranker developed by AdarshSingh7647. It is trained using the Std-SFT method, which optimizes the loss on the teacher's reasoning and final answer. This model functions as a pointwise reranker, similar to Rank1, by generating reasoning and then outputting 'true' or 'false' to indicate passage relevance. It is specifically designed for information retrieval tasks where ranking passages based on their relevance to a query is critical.

Loading preview...

Overview

AdarshSingh7647/Std-SFT-Passage-14B is a 14 billion parameter passage reranker built upon the Qwen3 architecture. Developed by AdarshSingh7647, this model utilizes a novel training approach called Std-SFT, where the loss function is applied to both the teacher's reasoning process and the final answer. This method aims to improve the model's ability to understand and articulate relevance.

Key Capabilities

  • Passage Reranking: Specializes in evaluating the relevance of a given passage to a specific query.
  • Pointwise Reranking: Operates by assessing each passage independently, generating internal reasoning before outputting a binary relevance decision ('true' or 'false').
  • Reasoning Generation: The model explicitly generates its reasoning process, ending with </think> true or </think> false, allowing for potential interpretability of its relevance judgments.
  • Qwen3 Base: Leverages the capabilities of the Qwen3 foundation model for robust language understanding.

Performance

While the Eklav version of this model (Eklav-Passage-14B) shows higher nDCG@10 scores on the BRIGHT benchmark, Std-SFT-Passage-14B achieves a BRIGHT nDCG@10 of 32.5, demonstrating its effectiveness in reranking tasks.

Good for

  • Information Retrieval Systems: Enhancing search result quality by re-ranking retrieved passages.
  • Question Answering Systems: Filtering and prioritizing relevant documents or snippets for answer generation.
  • Semantic Search: Improving the precision of search results by deeply understanding query-passage relationships.

Finetunes
1 models