dolev31

dolev31/TabAgent

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

2 stars 0 forks 2 watchers Python Apache License 2.0
ai-agents catboost distillation llm-agents machine-learning routing tabular-data tool-selection
View on GitHub Website
12 Open Issues Need Help Last updated: Oct 25, 2026

Open Issues Need Help

View All on GitHub
🧩 Track: a richer, clearer feature card 10 days ago
help wanted track: feature-card
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
🔁 Track: a smoother, safer hot swap 10 days ago
help wanted track: hot-swap
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
🔍 Track: find the LLM calls worth replacing, automatically 10 days ago
help wanted track: scan
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
A feature-card benchmark 10 days ago
help wanted track: feature-card research
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
`tabagent card explain`: see what a feature card holds and what the head uses 10 days ago
documentation help wanted track: feature-card
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
A new fitted primitive: candidate recency 10 days ago
good first issue track: feature-card
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
A cache-friendly tool list 10 days ago
help wanted track: hot-swap research
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
Shadow mode: watch the head before it takes over 10 days ago
help wanted track: hot-swap
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
A LangChain / LangGraph integration example 10 days ago
documentation good first issue track: hot-swap
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
A savings estimate in the scan report 10 days ago
help wanted track: scan
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
Read LangSmith and Langfuse exports 10 days ago
good first issue track: scan adapters
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection
Read OpenTelemetry GenAI traces in `tabagent scan` 10 days ago
good first issue track: scan adapters
dolev31/TabAgent
2

Replace an LLM agent's closed-set decisions with a cheap classifier trained from its own logs. Code for the paper TabAgent: Distilling LLM Agent Components into Cheap Classifiers.

Python
#ai-agents#catboost#distillation#llm-agents#machine-learning#routing#tabular-data#tool-selection