# Multi-task Generalization

GLM-4-9B-0414 is a lightweight member of the GLM family with 9 billion parameters, excelling in mathematical reasoning and general tasks, providing an efficient solution for resource-constrained scenarios.

Llama-3.1-8B model fine-tuned using the automatically filtered 50,000-entry Tulu-3-MIG dataset

An 8B-parameter language model based on the Llama 3.1 architecture, created by merging multiple specialized models, excelling in text generation tasks.

A 70B-parameter large language model merged using the TIES method, based on Steelskull/L3.3-MS-Nevoria-70b and fused with the DeepSeek-R1-Distill-Llama-70B model

A large language model based on Hermes-3-Llama-3.1-70B-lorablated, fine-tuned on the Gutenberg dataset

LWM is the first foundational model in the field of wireless communications, developed as a universal feature extractor capable of extracting fine-grained representations from wireless channel data.

RobustSAM is a model for robust segmentation of arbitrary objects in degraded images, improved upon SAM with enhanced segmentation performance on low-quality images.

Percival_01-7b-slerp is a 7B-parameter large language model ranked second on the OPENLLM leaderboard, obtained by merging the liminerity/M7-7b and Gille/StrangeMerges_32-7B-slerp models using the LazyMergekit tool.

TheBeagle is a 7-billion-parameter model trained on The Bagel dataset, optimized with DPO (Direct Preference Optimization) and UNA (Unified Neural Architecture) techniques, demonstrating excellent performance in multi-task scenarios.

This model is created by merging Locutusque/TinyMistral-248M-Instruct and Felladrin/TinyMistral-248M-SFT-v4 using the mergekit tool, combining the strengths of both. It possesses programming capabilities and reasoning skills while maintaining low hallucination and strong instruction-following abilities.

This is a hybrid model based on Llama-2-13B, incorporating features from multiple models including Tiefighter, MedLLaMA, and limarp-v2, suitable for various text generation tasks.

A reward model trained based on the LMFlow framework. It is trained on the HH - RLHF dataset (only the useful part) with open_llama_3b as the base model and has good generalization ability.