SecondLookResearch/teaching-qwen-why-data
Updated • 9
Teaching Qwen Why
Qwen 2.5 32B/72B LoRA adapters from Teaching Qwen Why: stories, instruct mix, difficult advice. Each card lists the chain to serve.
Note All training data, eval score tables and judge labels behind the post's figures.
SecondLookResearch/Qwen2.5-32B-instruct-mix
Updated • 42Note Figure 1: 32B baseline (instruct mix only).
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5
Updated • 22Note Figure 1: stories adapter (Sonnet 5 stories); serve with its instruct-mix adapter.
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-instruct-mix
Updated • 23Note Figure 1: 32B stories model.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-3000-1ep
Updated • 28Note Figure 1: 32B difficult advice (3,000 Terra conversations).
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-instruct-mix-difficult-advice-gpt-5.6-terra-3000-1ep
Updated • 31Note Figure 1: 32B stack (stories, instruct mix, difficult advice).
SecondLookResearch/Qwen2.5-72B-instruct-mix
Updated • 33Note Figure 1: 72B baseline.
SecondLookResearch/Qwen2.5-72B-stories-sonnet-5
Updated • 19Note Figure 1: 72B stories adapter.
SecondLookResearch/Qwen2.5-72B-stories-sonnet-5-instruct-mix
Updated • 31Note Figure 1: 72B stories model.
SecondLookResearch/Qwen2.5-72B-instruct-mix-difficult-advice-gpt-5.6-terra-1000-1ep
Updated • 31Note Figure 1: 72B difficult advice (1,000 Terra conversations).
SecondLookResearch/Qwen2.5-72B-stories-sonnet-5-instruct-mix-difficult-advice-gpt-5.6-terra-1000-1ep
Updated • 33Note Figure 1: 72B stack.
SecondLookResearch/Qwen2.5-32B-stories-haiku-4.5
UpdatedNote Figure 3: stories by Claude Haiku 4.5 (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-haiku-4.5-instruct-mix
Updated • 1Note Figure 3: stories model, Claude Haiku 4.5 teacher.
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano
Updated • 1Note Figure 3: stories by GPT-5.4 nano (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-instruct-mix
Updated • 2Note Figure 3: stories model, GPT-5.4 nano teacher.
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.6-terra
Updated • 1Note Figure 3: stories by GPT-5.6 Terra (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.6-terra-instruct-mix
Updated • 2Note Figure 3: stories model, GPT-5.6 Terra teacher.
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-recitation
Updated • 1Note Figure 4: recitation stories by Sonnet 5 (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-recitation-instruct-mix
Updated • 1Note Figure 4: recitation stories model, Sonnet 5.
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-recitation
Updated • 4Note Figure 4: recitation stories by GPT-5.4 nano (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-recitation-instruct-mix
Updated • 1Note Figure 4: recitation stories model, GPT-5.4 nano.
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-human
Updated • 4Note Stories rewritten with human protagonist (Figure 6); stories adapter.
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-human-instruct-mix
Updated • 3Note Stories model, human protagonist (Figure 6).
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-named-claude
Updated • 3Note Stories rewritten with Claude as protagonist (Figure 7); stories adapter.
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-named-claude-instruct-mix
Updated • 2Note Stories model, Claude as protagonist (Figure 7).
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-named-qwen
Updated • 4Note Stories rewritten with Qwen as protagonist (Figure 7); stories adapter.
SecondLookResearch/Qwen2.5-32B-stories-sonnet-5-named-qwen-instruct-mix
Updated • 4Note Stories model, Qwen as protagonist (Figure 7).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-3m
Updated • 3Note Figure 2: story ladder, 3M tokens (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-3m-instruct-mix
Updated • 1Note Figure 2: story ladder model, 3M tokens.
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-28m
Updated • 2Note Figure 2: story ladder, 28M tokens (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-28m-instruct-mix
Updated • 2Note Figure 2: story ladder model, 28M tokens.
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-56m
Updated • 1Note Figure 2: story ladder, 56M tokens (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-56m-instruct-mix
Updated • 3Note Figure 2: story ladder model, 56M tokens.
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-112m
Updated • 1Note Figure 2: story ladder, 112M tokens (stories adapter).
SecondLookResearch/Qwen2.5-32B-stories-gpt-5.4-nano-112m-instruct-mix
Updated • 1Note Figure 2: story ladder model, 112M tokens.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-500-1ep
Updated • 23Note Figure 2: difficult advice, 500 Terra conversations.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-1000-1ep
Updated • 30Note Figure 2 and 3: difficult advice, 1,000 Terra conversations.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-2000-1ep
Updated • 21Note Figure 2: difficult advice, 2,000 Terra conversations.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-sonnet-5-1000-1ep
Updated • 37Note Figure 3: difficult advice, 1,000 Sonnet 5 conversations.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-135-10ep
Updated • 1Note Figures 12 and 14: Terra prompts and responses, 135 conversations, 10 epochs.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-sonnet-5-gpt-5.6-terra-prompts-135-10ep
Updated • 1Note Figure 12: Sonnet 5 responses to Terra prompts.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-sonnet-5-135-10ep
Updated • 22Note Figure 12: Sonnet 5 prompts and responses.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-sonnet-5-prompts-135-10ep
Updated • 18Note Figure 12: Terra responses to Sonnet 5 prompts.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-135-20ep
Updated • 3Note Figure 14: 135 Terra conversations, 20 epochs.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-285-5ep
Updated • 2Note Figure 14: 285 Terra conversations, 5 epochs.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-285-10ep
Updated • 24Note Figure 14: 285 Terra conversations, 10 epochs.
SecondLookResearch/Qwen2.5-32B-instruct-mix-difficult-advice-gpt-5.6-terra-285-20ep
Updated • 29Note Figure 14: 285 Terra conversations, 20 epochs.