Update model card

README.md

@@ -12,13 +12,25 @@ tags:
 
 # TabPilot
 
-Trained agents of [TabPilot](https://github.com/fearlessAAA/TabPilot), which answers
-natural-language tabular prediction requests with two Qwen3.5-4B agents and a frozen tabular
-foundation model (TabPFN v3.5).
+**TabPilot: Empowering Small Language Models for Natural-Language Tabular Prediction via
+Declarative Query Compilation, Privileged Distillation, and Structure-Aware Rewards**
+· Code: [github.com/fearlessAAA/TabPilot](https://github.com/fearlessAAA/TabPilot)
+
+TabPilot is an agentic framework that couples a 4B small language model with a **plug-and-play
+frozen tabular foundation model (TFM)** through **declarative predictive query compilation**. An
+**Execution Agent** iteratively compiles each natural-language request into a structured
+prediction query and validates it through leakage-safe tools; the frozen TFM executes the query
+by in-context inference; a **Response Agent** synthesizes the TFM predictions into a grounded
+answer. The agents are trained for unseen tables with a three-stage local-to-global curriculum:
+**Privileged Semantic On-Policy Self-Distillation (PS-OPSD)**, **Structure-Aware GRPO (SA-GRPO)**
+with an outcome-label-free, TFM-agnostic structural reward, and **Teacher-Guided Tabular Response
+Distillation (TTRD)**.
+
+This repository holds the trained agents (backbone: Qwen3.5-4B, non-thinking mode):
 
 | Folder | Model |
 |---|---|
-| `execution-agent/fold<k>/` | Execution Agent: full Qwen3.5-4B weights after PS-OPSD and SA-GRPO |
+| `execution-agent/fold<k>/` | Execution Agent: full weights after PS-OPSD → SA-GRPO |
 | `response-agent/fold<k>/` | Response Agent: TTRD LoRA adapter (rank 32) for `Qwen/Qwen3.5-4B` |
 
 There is one model pair per fold `k` of the five-fold cross-validation over TopBench source tables
@@ -31,8 +43,9 @@ tables, so evaluate it only on fold `k`'s test tables.
 hf download XiaoboX/TabPilot --include "execution-agent/fold0/*" "response-agent/fold0/*" --local-dir checkpoints
 ```
 
-Serve both agents with vLLM and run TabPilot as described in the code repository's README. Both
-agents run in non-thinking mode.
+Serve both agents with vLLM and run TabPilot as described in the code repository's README. The
+TFM is plug-and-play: the agents were trained without calling it, and the code repository uses
+TabPFN v3.5 by default.
 
 ## License