Language Models Comparison Hub
18 papers - avg viability 4.6
Recent advancements in language models are increasingly focused on enhancing accessibility and efficiency, particularly for low-resource languages. Innovations like Kakugo enable the creation of small language models for 54 languages at minimal cost, democratizing AI development for underserved communities. Meanwhile, techniques such as reward-guided stitching in diffusion models are improving reasoning capabilities by aggregating intermediate outputs, leading to significant accuracy gains in complex tasks. The introduction of specialized models like LilMoo for Hindi and Sabiá-4 for Brazilian Portuguese highlights a trend toward tailored solutions that outperform larger multilingual counterparts in specific linguistic contexts. Additionally, value-aware numerical representations are addressing fundamental weaknesses in numerical reasoning, while low-resolution visual tokens are being explored to enrich character modeling in languages like Chinese. Collectively, these efforts are reshaping the landscape of language modeling, making it more inclusive and robust for diverse applications across various languages and tasks.
Top Papers
- Kakugo: Distillation of Low-Resource Languages into Small Language Models(8.0)
Kakugo: Cost-effective pipeline for developing AI models in low-resource languages using distillation under $50 per language.
- Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching(8.0)
Develop a reward-guided diffusion language model tool for enhanced reasoning across math and coding tasks.
- A.X K1 Technical Report(6.0)
A.X K1 is a scalable MoE language model excelling in Korean-language benchmarks with user-controlled reasoning capabilities.
- Sabiá-4 Technical Report(6.0)
Sabiá-4 is a new generation of Portuguese language models optimized for legal and conversational tasks.
- Value-Aware Numerical Representations for Transformer Language Models(6.0)
Enhance language models to handle numerical data robustly with value-aware token embeddings.
- Hot-Start from Pixels: Low-Resolution Visual Tokens for Chinese Language Modeling(6.0)
Develop a Chinese language model using low-resolution character images for improved efficiency and prediction.
- Raising Bars, Not Parameters: LilMoo Compact Language Model for Hindi(6.0)
LilMoo offers a compact, high-performance language model tailored specifically for Hindi, enabling more equitable NLP innovations.
- Parallelism and Generation Order in Masked Diffusion Language Models: Limits Today, Potential Tomorrow(5.0)
Explore parallel token generation in language models with our innovative diffusion approach.
- Vision-DeepResearch: Incentivizing DeepResearch Capability in Multimodal Large Language Models(5.0)
Vision-DeepResearch enhances multimodal large language models with advanced deep-research capabilities for improved real-world search performance.
- Language Model Inversion through End-to-End Differentiation(5.0)
Optimizing input prompts for language model outputs using end-to-end differentiation.