Update README.md

README.md

@@ -13,7 +13,7 @@ datasets:
 - svamp
 - multi_arith
 model-index:
-- name: SIM_COT-LLaMA3-CODI-8B
+- name: SIM_COT-LLaMA3-CODI-1B
   results:
   - task:
       type: math-word-problems
@@ -44,7 +44,7 @@ model-index:
       value: xx.x
 ---
 
-# 🚀 SIM_COT-LLaMA3-CODI-8B
+# 🚀 SIM_COT-LLaMA3-CODI-1B
 
 [![🤗 Model Repo](https://img.shields.io/badge/HuggingFace-Model-blue)](https://huggingface.co/internlm/SIM_COT-LLaMA3-CODI-8B)
 [![📂 GitHub](https://img.shields.io/badge/Code-GitHub-black?logo=github)](https://github.com/InternLM/SIM-CoT)  
@@ -66,7 +66,7 @@ Empirical results demonstrate that SIM-CoT substantially improves both **in-doma
 
 ---
 
-**SIM_COT-LLaMA3-CODI-8B** is a large implicit language model based on **Meta LLaMA-3.1-8B-Instruct**, fine-tuned with **SIM-CoT (Supervised Implicit Chain-of-Thought)** on top of the **CODI latent reasoning framework**.  
+**SIM_COT-LLaMA3-CODI-1B** is a large implicit language model based on **Meta LLaMA-3.2-1B-Instruct**, fine-tuned with **SIM-CoT (Supervised Implicit Chain-of-Thought)** on top of the **CODI latent reasoning framework**.  
 It is designed to improve ✨ *implicit reasoning* and 🧮 *arithmetic multi-step problem solving* across benchmarks such as **GSM8K, GSM-Hard, MultiArith, and SVAMP**.
 
 ---
@@ -98,9 +98,9 @@ We evaluate **SIM-CoT** across both **in-domain** (GSM8K-Aug) and **out-of-domai
 
 ## 📌 Model Details
 
-- 🏗️ **Base model**: [LLaMA-3.1-8B-Instruct](https://huggingface.co/meta-llama/Llama-3.1-8B-Instruct)  
+- 🏗️ **Base model**: [LLaMA-3.2-1B-Instruct](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct)  
 - ⚡ **Fine-tuning method**: LoRA (r=128, alpha=32)  
-- 🔑 **Latent reasoning**: 6 latent steps, projection dimension = 4096  
+- 🔑 **Latent reasoning**: 6 latent steps, projection dimension = 2048  
 - 🎯 **Dropout**: 0.0 (projection layer)  
 - 🖥️ **Precision**: bf16  
 - 📏 **Context length**: 512 tokens  
@@ -132,32 +132,32 @@ cd SIM-CoT/CODI
 
 ### 2. Run the evaluation script
 We provide shell scripts for different backbones and datasets.
-For example, to evaluate on **LLaMA-3.1 8B** with the **SVAMP** dataset, run:
+For example, to evaluate on **LLaMA-3.2 1B** with the **SVAMP** dataset, run:
 ```
-bash test_llama8b.sh
+bash test_llama1b.sh
 ```
 This will internally call the following command:
 ```
 python test.py \
-  --data_name "svamp" \
-  --output_dir "$SAVE_DIR" \
-  --model_name_or_path path/to/Llama-3.1-8B-Instruct \
-  --seed 11 \
-  --model_max_length 512 \
-  --bf16 \
-  --lora_r 128 --lora_alpha 32 --lora_init \
-  --batch_size 128 \
-  --greedy True \
-  --num_latent 6 \
-  --use_prj True \
-  --prj_dim 4096 \
-  --prj_no_ln False \
-  --prj_dropout 0.0 \
-  --inf_latent_iterations 6 \
-  --inf_num_iterations 1 \
-  --remove_eos True \
-  --use_lora True \
-  --ckpt_dir path/to/sim_cot-checkpoints
+	--data_name "svamp" \
+	--output_dir "$SAVE_DIR" \
+	--model_name_or_path path/to/Llama-3.2-1B-Instruct \
+	--seed 11 \
+	--model_max_length 512 \
+	--bf16 \
+	--lora_r 128 --lora_alpha 32 --lora_init \
+	--batch_size 128 \
+	--greedy True \
+	--num_latent 6 \
+	--use_prj True \
+	--prj_dim 2048 \
+	--prj_no_ln False \
+	--prj_dropout 0.0 \
+	--inf_latent_iterations 6 \
+	--inf_num_iterations 1 \
+	--remove_eos True \
+	--use_lora True \
+	--ckpt_dir path/to/sim_cot-checkpoints
 ```
 ### 3. Expected output
 After running, the script will print the evaluation summary.