laion/100k_baseline__Qwen3-8B
The laion/100k_baseline__Qwen3-8B is an 8 billion parameter language model, fine-tuned from the Qwen/Qwen3-8B architecture. It has a context length of 32768 tokens and is specifically trained on a diverse collection of agent-based interaction traces. This model is optimized for tasks requiring complex reasoning and decision-making within simulated environments, making it suitable for agentic applications.
Loading preview...
Overview
This model, laion/100k_baseline__Qwen3-8B, is an 8 billion parameter language model derived from the Qwen3-8B architecture. It features a substantial context length of 32768 tokens, enabling it to process and understand extensive inputs. The model has undergone fine-tuning on a specialized collection of datasets, primarily consisting of agent-based interaction traces from various simulated environments. These datasets include traces from swesmith-sandboxes-with_tests, r2egym experiments (including askllm-hardened and constrained variants), tas_optimal_combined_traces, and Kimi-K2T-swesmith.
Key Capabilities
- Agentic Reasoning: Fine-tuned on diverse agent interaction traces, suggesting capabilities in understanding and generating responses for complex, multi-step tasks within simulated environments.
- Extended Context Handling: With a 32768-token context window, it can process and maintain coherence over long dialogues or detailed problem descriptions.
Training Details
The model was trained with a learning rate of 4e-05 over 7 epochs, utilizing a distributed setup across 128 devices. The training employed an AdamW optimizer with specific beta and epsilon parameters, and a cosine learning rate scheduler with a 0.1 warmup ratio.
Good For
- Developing and evaluating AI agents in simulated environments.
- Tasks requiring understanding of complex interaction logs and decision-making processes.
- Applications benefiting from a large context window for detailed problem-solving.