cacherepair is an open-weight model from Genglin Wang, released under other. Its published files total 2.3 GB.
Repair networks for CacheRepair: Learning to Repair Cross-Chunk Context in RAG for KV Cache Fusion. Each network predicts residual corrections to independently precomputed document KV caches for one frozen target LLM.
Model Card
Repair networks for CacheRepair: Learning to Repair Cross-Chunk Context in RAG for KV Cache Fusion. Each network predicts residual corrections to independently precomputed document KV caches for one frozen target LLM. The same repairer serves MuSiQue, HotpotQA, MultiHop-RAG, and TriviaQA. The Qwen repairers are Built with Qwen. The Llama repairers are Built with Llama and are named Llama-CacheRepair S/M/L. Each target/size directory contains model.safetensors and config.json. The weight file holds repair parameters and the sigmastale / sigmadelta normalization tensors in FP32. The target LLM supplies its own frozen embedding. manifest.json lists exact parameter counts and file identities.…
Excerpt from the card by Genglin Wang, licensed other.
Identity and Version
- Repository
- gwang3456/cacherepair
- Publisher
- Genglin Wang
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- en
- Revision
- 0f96dbf4b12abe2d634a646fb3273b226a41fd67
- First published
- 2026-08-20
- Last updated
- 2026-09-29
Files and Weights
27 files, 2.3 GB in total. The weights are 9 files totalling 2.3 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| Llama-3.1-8B-Instruct/L/model.safetensors | Weights | 372.8 MB | 995f57e3abda |
| Llama-3.1-8B-Instruct/M/model.safetensors | Weights | 235.9 MB | 91fe41571637 |
| Llama-3.1-8B-Instruct/S/model.safetensors | Weights | 93.4 MB | c424197d08c9 |
| Qwen2.5-14B-Instruct/L/model.safetensors | Weights | 635.3 MB | 5f7203250a67 |
| Qwen2.5-14B-Instruct/M/model.safetensors | Weights | 415.3 MB | 542d8990c557 |
| Qwen2.5-14B-Instruct/S/model.safetensors | Weights | 168.0 MB | 6710f1ad6305 |
| Qwen2.5-3B-Instruct/L/model.safetensors | Weights | 204.4 MB | 86e5b06c8c48 |
| Qwen2.5-3B-Instruct/M/model.safetensors | Weights | 119.5 MB | cbe900e8829b |
| Qwen2.5-3B-Instruct/S/model.safetensors | Weights | 38.0 MB | 5a9f38593146 |
| Llama-3.1-8B-Instruct/L/config.json | Configuration | 2.1 KB | — |
| Llama-3.1-8B-Instruct/M/config.json | Configuration | 2.1 KB | — |
| Llama-3.1-8B-Instruct/S/config.json | Configuration | 2.1 KB | — |
| Qwen2.5-14B-Instruct/L/config.json | Configuration | 1.7 KB | — |
| Qwen2.5-14B-Instruct/M/config.json | Configuration | 1.7 KB | — |
| Qwen2.5-14B-Instruct/S/config.json | Configuration | 1.7 KB | — |
| Qwen2.5-3B-Instruct/L/config.json | Configuration | 1.7 KB | — |
| Qwen2.5-3B-Instruct/M/config.json | Configuration | 1.7 KB | — |
| Qwen2.5-3B-Instruct/S/config.json | Configuration | 1.7 KB | — |
| manifest.json | Configuration | 3.5 KB | — |
| LICENSE | Documentation | 895 B | — |
| Llama-3.1-8B-Instruct/LICENSE | Documentation | 7.7 KB | — |
| Llama-3.1-8B-Instruct/USE_POLICY.md | Documentation | 4.7 KB | — |
| NOTICE | Documentation | 711 B | — |
| Qwen2.5-14B-Instruct/LICENSE | Documentation | 11.4 KB | — |
| Qwen2.5-3B-Instruct/LICENSE | Documentation | 7.4 KB | — |
| README.md | Documentation | 4.5 KB | — |
| .gitattributes | Repository | 50 B | — |
License and Download
- License
- other
- Access
- Open weights, no gate
- Download size
- 2.3 GB
Released by Genglin Wang through its official repository on Hugging Face.
Built From
- Derived from Qwen/Qwen2.5-14B-Instruct
- Described by arXiv:2609.35139
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 2.3 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About cacherepair
What license is cacherepair released under?
other, as its publisher declares it. Read the license text before commercial use.