Tool A/B Spot Check - Base vs Tuned (2026-05-30)

Generated: 2026-05-30 23:49:18Z

Scope

• Manual spot-check battery: 5 prompts per tool across 12 Athena tools (60 total A/B pairs)
• Base model: Qwen/Qwen3.5-4B
• Tuned adapter: D:\temp\memorysmith-training\runs\distilled-all-cat-20260530-v6-chatml\adapter
• Raw results JSON: Data/Pages/research/training/tool-ab-spotcheck-20260530-v5.data.json

Headline Metrics

Per-Tool Results

Notable Improvements

• memorysmith_unified_search-1 (memorysmith_unified_search): base(match=False, env=False, pred=None) -> tuned(match=False, env=True, pred=memorysmith_search); prompt="search the wiki for kv cache options"
• memorysmith_unified_search-2 (memorysmith_unified_search): base(match=False, env=False, pred=None) -> tuned(match=False, env=True, pred=memorysmith_code_search); prompt="find docs for training harness"
• memorysmith_unified_search-3 (memorysmith_unified_search): base(match=False, env=False, pred=None) -> tuned(match=False, env=True, pred=memorysmith_code_search); prompt="lookup wiki notes about chat template"
• memorysmith_unified_search-4 (memorysmith_unified_search): base(match=False, env=False, pred=None) -> tuned(match=False, env=True, pred=memorysmith_code_search); prompt="search for model profile defaults"
• memorysmith_unified_search-5 (memorysmith_unified_search): base(match=False, env=False, pred=None) -> tuned(match=False, env=True, pred=memorysmith_code_search); prompt="find references about code search eta"
• memorysmith_hybrid_search-1 (memorysmith_hybrid_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_hybrid_search); prompt="run a hybrid search for request guard middleware"
• memorysmith_hybrid_search-2 (memorysmith_hybrid_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_hybrid_search); prompt="hybrid search for source bundle auth behavior"
• memorysmith_hybrid_search-3 (memorysmith_hybrid_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_hybrid_search); prompt="use hybrid search to find onnx semantic search notes"
• memorysmith_hybrid_search-4 (memorysmith_hybrid_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_hybrid_search); prompt="hybrid search for maintenance proposal workflow"
• memorysmith_hybrid_search-5 (memorysmith_hybrid_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_hybrid_search); prompt="run hybrid search for task assignee update invariant"
• memorysmith_semantic_search-1 (memorysmith_semantic_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_semantic_search); prompt="semantic search for retrieval warning propagation"
• memorysmith_semantic_search-2 (memorysmith_semantic_search): base(match=False, env=False, pred=None) -> tuned(match=True, env=True, pred=memorysmith_semantic_search); prompt="semantic search for pages lock navigation regression"

Notable Regressions

• None

Persistent Failures (Both Models)

• memorysmith_unified_search-1 (memorysmith_unified_search): base pred=None, tuned pred=memorysmith_search, baseErr=No JSON object found, tunedErr=None
• memorysmith_unified_search-2 (memorysmith_unified_search): base pred=None, tuned pred=memorysmith_code_search, baseErr=No JSON object found, tunedErr=None
• memorysmith_unified_search-3 (memorysmith_unified_search): base pred=None, tuned pred=memorysmith_code_search, baseErr=No JSON object found, tunedErr=None
• memorysmith_unified_search-4 (memorysmith_unified_search): base pred=None, tuned pred=memorysmith_code_search, baseErr=No JSON object found, tunedErr=None
• memorysmith_unified_search-5 (memorysmith_unified_search): base pred=None, tuned pred=memorysmith_code_search, baseErr=No JSON object found, tunedErr=None
• memorysmith_search-3 (memorysmith_search): base pred=None, tuned pred=memorysmith_text_search, baseErr=No JSON object found, tunedErr=None
• memorysmith_get-2 (memorysmith_get): base pred=None, tuned pred=memorysmith_open, baseErr=No JSON object found, tunedErr=None
• memorysmith_get-5 (memorysmith_get): base pred=None, tuned pred=memorysmith_load, baseErr=No JSON object found, tunedErr=None
• memorysmith_page_search-3 (memorysmith_page_search): base pred=None, tuned pred=memorysmith_page_notes, baseErr=No JSON object found, tunedErr=None
• memorysmith_page_get-1 (memorysmith_page_get): base pred=None, tuned pred=memorysmith_page_open, baseErr=No JSON object found, tunedErr=None
• memorysmith_page_get-3 (memorysmith_page_get): base pred=None, tuned pred=memorysmith_page, baseErr=No JSON object found, tunedErr=None
• memorysmith_page_get-5 (memorysmith_page_get): base pred=None, tuned pred=memorysmith_page_open, baseErr=No JSON object found, tunedErr=None
• memorysmith_task_get-2 (memorysmith_task_get): base pred=None, tuned pred=memorysmith_task_open, baseErr=No JSON object found, tunedErr=None
• memorysmith_task_get-5 (memorysmith_task_get): base pred=None, tuned pred=memorysmith_task_load, baseErr=No JSON object found, tunedErr=None
• memorysmith_code_search-4 (memorysmith_code_search): base pred=None, tuned pred=memorysmith_code, baseErr=No JSON object found, tunedErr=None

Representative Output Snippets

memorysmith_unified_search-1

• Prompt: search the wiki for kv cache options
• Base: [think] Okay, the user is asking me to search the wiki for information about kv cache options. I need to remember that I'm Athena, MemorySmith's local wiki assistant. When a search or retrieval action is requested, I should respond with ...
• Tuned: [think] {"toolCalls":[{"name":"memorysmith_search","arguments":{"query":"kv cache options"}}]} assistant [think] {"toolCalls":[{"name":"memorysmith_search","arguments":{"query":"kv cache options"}}]} assistant [think] {"toolCalls":[{"nam...

memorysmith_unified_search-2

• Prompt: find docs for training harness
• Base: assistant [think] Okay, the user is asking for documentation on the training harness. I need to figure out how to respond as Athena, the local wiki assistant. First, I should check if there's a specific tool I need to use for retrieving ...
• Tuned: assistant [think] {"toolCalls":[{"name":"memorysmith_code_search","arguments":{"query":"training harness","limit":5}}]}

memorysmith_unified_search-3

• Prompt: lookup wiki notes about chat template
• Base: assistant [think] Okay, the user is asking me to look up wiki notes about chat templates. I need to check if there's any information in the MemorySmith's wiki about chat templates. Since I'm Athena, the local wiki assistant, I should use...
• Tuned: assistant [think] {"toolCalls":[{"name":"memorysmith_code_search","arguments":{"query":"chat template","limit":5}}]} assistant [think] {"toolCalls":[{"name":"memorysmith_code_search","arguments":{"query":"chat template","limit":5}}]} ass...

memorysmith_unified_search-4

• Prompt: search for model profile defaults
• Base: assistant [think] Okay, the user is asking me to search for model profile defaults. I need to figure out what they mean by that. Model profiles could refer to different things depending on the context. Maybe they're talking about machine...
• Tuned: assistant [think] {"toolCalls":[{"name":"memorysmith_code_search","arguments":{"query":"model profile defaults","limit":10}}]}

memorysmith_unified_search-5

• Prompt: find references about code search eta
• Base: assistant [think] Okay, the user is asking for references about code search ETA. First, I need to figure out what they mean by "code search eta". ETA usually stands for Estimated Time of Arrival, but in the context of code search, that d...
• Tuned: assistant [think] {"toolCalls":[{"name":"memorysmith_code_search","arguments":{"query":"eta","limit":10}}]}

memorysmith_hybrid_search-1

• Prompt: run a hybrid search for request guard middleware
• Base: [think] Okay, the user is asking me to run a hybrid search for "request guard middleware". I need to figure out how to handle this. First, I remember that as Athena, I'm supposed to respond with a JSON object when a search/retrieval acti...
• Tuned: [think] {"toolCalls":[{"name":"memorysmith_hybrid_search","arguments":{"query":"request guard middleware","memoryLimit":6}}]} assistant [think] {"toolCalls":[{"name":"memorysmith_hybrid_search","arguments":{"query":"request guard middlew...

memorysmith_hybrid_search-2

• Prompt: hybrid search for source bundle auth behavior
• Base: [think] Okay, the user is asking about hybrid search for source bundle auth behavior. Let me start by recalling what I know about hybrid search. Hybrid search usually combines different search methods, like keyword and vector search, to ...
• Tuned: [think] {"toolCalls":[{"name":"memorysmith_hybrid_search","arguments":{"query":"source bundle auth behavior","limit":6}}]} assistant [think] {"toolCalls":[{"name":"memorysmith_hybrid_search","arguments":{"query":"source bundle auth behav...

memorysmith_hybrid_search-3

• Prompt: use hybrid search to find onnx semantic search notes
• Base: Okay, the user is asking me to use hybrid search to find ONNX semantic search notes. First, I need to understand what they're looking for. Hybrid search typically combines different search methods, like keyword and semantic search, to re...
• Tuned: {"toolCalls":[{"name":"memorysmith_hybrid_search","arguments":{"query":"onnx","memoryLimit":10}}]} assistant [think] {"toolCalls":[{"name":"memorysmith_hybrid_search","arguments":{"query":"onnx","memoryLimit":10}}]} assistant