diff --git a/.github/workflows/docker/docker-compose.yaml b/.github/workflows/docker/docker-compose.yaml
index 5fe4571b57..3ccb526b75 100644
--- a/.github/workflows/docker/docker-compose.yaml
+++ b/.github/workflows/docker/docker-compose.yaml
@@ -6,12 +6,12 @@ services:
     environment:
       - HF_ENDPOINT=https://hf-mirror.com
       - RAY_ADDRESS=auto
-      - CHECKPOINT_ROOT_DIR=/mnt/checkpoints
-      - DATA_ROOT_DIR=/mnt/data
-      - MODEL_PATH=/mnt/models/Qwen3-0.6B
-      - API_MODEL_PATH=/mnt/models/Qwen3-1.7B
-      - VLM_MODEL_PATH=/mnt/models/Qwen2.5-VL-3B
-      - CHECKPOINT_PATH=/mnt/checkpoints
+      - TRINITY_CHECKPOINT_ROOT_DIR=/mnt/checkpoints
+      - TRINITY_TASKSET_PATH=/mnt/data
+      - TRINITY_SFT_DATASET_PATH=/mnt/data
+      - TRINITY_MODEL_PATH=/mnt/models/Qwen3-0.6B
+      - TRINITY_API_MODEL_PATH=/mnt/models/Qwen3-1.7B
+      - TRINITY_VLM_MODEL_PATH=/mnt/models/Qwen2.5-VL-3B
     working_dir: /workspace
     networks:
       - trinity-network
@@ -33,10 +33,9 @@ services:
     command: sh -c "pip install -e .[dev] && ray start --address=trinity-node-1:6379 --block"
     environment:
       - HF_ENDPOINT=https://hf-mirror.com
-      - CHECKPOINT_ROOT_DIR=/mnt/checkpoints
-      - DATA_ROOT_DIR=/mnt/data
-      - MODEL_PATH=/mnt/models/Qwen3-1.7B
-      - CHECKPOINT_PATH=/mnt/checkpoints
+      - TRINITY_CHECKPOINT_ROOT_DIR=/mnt/checkpoints
+      - TRINITY_TASKSET_PATH=/mnt/data
+      - TRINITY_MODEL_PATH=/mnt/models/Qwen3-1.7B
     working_dir: /workspace
     volumes:
       - trinity-volume:/mnt
diff --git a/benchmark/bench.py b/benchmark/bench.py
index 1aafd2b55c..9d7c634527 100644
--- a/benchmark/bench.py
+++ b/benchmark/bench.py
@@ -8,6 +8,7 @@
 import yaml
 
 from trinity.algorithm.algorithm import ALGORITHM_TYPE
+from trinity.common.constants import MODEL_PATH_ENV_VAR
 from trinity.utils.dlc_utils import get_dlc_env_vars
 
 
@@ -77,9 +78,8 @@ def prepare_configs(args, rank, current_time):
         set_engine_num(config, args)
         config["model"]["model_path"] = (
             args.model_path
-            or os.environ.get("MODEL_PATH")
             or config["model"]["model_path"]
-            or "Qwen/Qwen2.5-1.5B-Instruct"
+            or os.environ.get(MODEL_PATH_ENV_VAR, "Qwen/Qwen2.5-1.5B-Instruct")
         )
         if ALGORITHM_TYPE.get(config["algorithm"]["algorithm_type"]).use_critic:
             config["model"]["critic_model_path"] = (
diff --git a/docs/sphinx_doc/source/tutorial/example_async_mode.md b/docs/sphinx_doc/source/tutorial/example_async_mode.md
index 3669271207..bf1015db29 100644
--- a/docs/sphinx_doc/source/tutorial/example_async_mode.md
+++ b/docs/sphinx_doc/source/tutorial/example_async_mode.md
@@ -15,12 +15,12 @@ Assuming we have a node with 8 GPUs, we allocate 4 GPUs for the trainer and 4 GP
 project: <project_name>
 name: <experiment_name>
 mode: explore
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
 cluster:
   node_num: 1
   gpu_per_node: 4
@@ -31,7 +31,8 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
       split: train
       format:
         prompt_key: 'question'
@@ -61,12 +62,12 @@ Key configurations in `trainer.yaml` are as follows:
 project: <project_name>
 name: <experiment_name>
 mode: train
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
 cluster:
   node_num: 1
   gpu_per_node: 4
@@ -77,7 +78,8 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
       format:
         prompt_key: 'question'
         response_key: 'answer'
@@ -129,12 +131,12 @@ Trinity-RFT also supports dynamic scaling in asynchronous mode. Continuing with
 project: <project_name>
 name: <experiment_name>
 mode: explore
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
 cluster:  # important
   node_num: 1
   gpu_per_node: 8
@@ -151,7 +153,8 @@ buffer:
     taskset:  # important
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
       format:
         prompt_key: 'question'
         response_key: 'answer'
diff --git a/docs/sphinx_doc/source/tutorial/example_data_functionalities.md b/docs/sphinx_doc/source/tutorial/example_data_functionalities.md
index 73fe56c729..7f62add78a 100644
--- a/docs/sphinx_doc/source/tutorial/example_data_functionalities.md
+++ b/docs/sphinx_doc/source/tutorial/example_data_functionalities.md
@@ -53,7 +53,7 @@ data_processor:
           input_keys: ["question", "answer"]
           field_names: ["Question", "Answer"]
     inputs:  # the output will be set to the explorer input automatically
-      - /PATH/TO/GSM8K/DATA/FILE
+      - ${oc.env:TRINITY_TASKSET_PATH}
     target_fields: ["question", "answer"]
 service:
   data_juicer:
diff --git a/docs/sphinx_doc/source/tutorial/example_dpo.md b/docs/sphinx_doc/source/tutorial/example_dpo.md
index 2841690415..3b50ed58c0 100644
--- a/docs/sphinx_doc/source/tutorial/example_dpo.md
+++ b/docs/sphinx_doc/source/tutorial/example_dpo.md
@@ -63,9 +63,9 @@ algorithm:
   kl_loss_fn: k1
   kl_loss_fn_args:
     kl_coef: 0.1  # value of beta in DPO
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: $MODEL_PATH/Qwen2.5-1.5B-Instruct
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
 cluster:
   node_num: 1
   gpu_per_node: 8
@@ -111,9 +111,9 @@ name: <experiment_name>
 mode: train
 algorithm:
   algorithm_type: sft
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
 cluster:
   node_num: 1
   gpu_per_node: 2
diff --git a/docs/sphinx_doc/source/tutorial/example_reasoning_basic.md b/docs/sphinx_doc/source/tutorial/example_reasoning_basic.md
index 01c69eb1e0..048842bd50 100644
--- a/docs/sphinx_doc/source/tutorial/example_reasoning_basic.md
+++ b/docs/sphinx_doc/source/tutorial/example_reasoning_basic.md
@@ -106,12 +106,12 @@ We use the configurations in [`gsm8k.yaml`](https://github.com/modelscope/Trinit
 ```yaml
 project: <project_name>
 name: <experiment_name>
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
 cluster:
   node_num: 1
   gpu_per_node: 2
@@ -122,7 +122,7 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: <$DATASET_PATH/gsm8k>
+      path: 'openai/gsm8k'
       subset_name: 'main'
       split: 'train'
       format:
@@ -133,7 +133,7 @@ buffer:
     eval_tasksets:
     - name: gsm8k-eval
       storage_type: file
-      path: <$DATASET_PATH/gsm8k>
+      path: 'openai/gsm8k'
       subset_name: 'main'
       split: 'test'
       format:
diff --git a/docs/sphinx_doc/source/tutorial/example_step_wise.md b/docs/sphinx_doc/source/tutorial/example_step_wise.md
index cbf45dfe09..73d6042de2 100644
--- a/docs/sphinx_doc/source/tutorial/example_step_wise.md
+++ b/docs/sphinx_doc/source/tutorial/example_step_wise.md
@@ -91,13 +91,13 @@ The example configuration is shown as:
 ```yaml
 project: "ALFWORLD"
 name: "Step_Wise_Alfworld"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/ALFWORLD_RFT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 16
   advantage_fn: step_wise_grpo
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 16384
   max_model_len: 20480
 cluster:
@@ -141,7 +141,7 @@ explorer:
     gpu_memory_utilization: 0.7
     enable_chunked_prefill: true
   env_vars:
-    TMPDIR: /PATH/TO/ALFWORLD_TMP_DIR
+    TMPDIR: ${oc.env:TMPDIR,/tmp}
 synchronizer:
   sync_style: dynamic_by_explorer
   sync_method: 'nccl'
diff --git a/docs/sphinx_doc/source/tutorial/trinity_configs.md b/docs/sphinx_doc/source/tutorial/trinity_configs.md
index d2ac4a10d1..0bafd44be7 100644
--- a/docs/sphinx_doc/source/tutorial/trinity_configs.md
+++ b/docs/sphinx_doc/source/tutorial/trinity_configs.md
@@ -11,7 +11,7 @@ The configuration for **Trinity-RFT** is defined in a `YAML` file and organized
 project: Trinity-RFT
 name: example
 mode: both
-checkpoint_root_dir: /PATH/TO/CHECKPOINT
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 continue_from_checkpoint: true
 
 algorithm:
@@ -67,7 +67,7 @@ These are general settings that apply to the entire experiment.
 project: Trinity-RFT
 name: example
 mode: both
-checkpoint_root_dir: ${oc.env:CHECKPOINT_ROOT_DIR}   # CHECKPOINT_ROOT_DIR is an environment variable set in advance
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}   # TRINITY_CHECKPOINT_ROOT_DIR is an environment variable set in advance
 ```
 
 - `project`: The name of the project.
@@ -214,7 +214,7 @@ buffer:
     taskset:
       name: countdown_train
       storage_type: file
-      path: /PATH/TO/DATA
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       split: train
       format:
         prompt_key: 'question'
@@ -227,7 +227,7 @@ buffer:
     eval_tasksets:
     - name: countdown_eval
       storage_type: file
-      path: /PATH/TO/DATA
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       split: test
       repeat_times: 1
       format:
@@ -281,7 +281,7 @@ buffer:
     sft_warmup_dataset:
       name: warmup_data
       storage_type: file
-      path: /PATH/TO/WARMUP_DATA
+      path: ${oc.env:TRINITY_SFT_DATASET_PATH}
       format:
         prompt_key: 'question'
         response_key: 'answer'
@@ -338,7 +338,7 @@ explorer:
     tensor_parallel_size: 1
     enable_history: False
   auxiliary_models:
-  - model_path: /PATH/TO/MODEL
+  - model_path: Qwen/Qwen2.5-7B-Instruct
     tensor_parallel_size: 1
   eval_interval: 100
   eval_on_startup: True
@@ -438,7 +438,7 @@ data_processor:
           input_keys: ["question", "answer"]
           field_names: ["Question", "Answer"]
     inputs:  # the output will be set to the explorer input automatically
-      - /PATH/TO/GSM8K/DATA/FILE
+      - ${oc.env:TRINITY_TASKSET_PATH}
     target_fields: ["question", "answer"]
   experience_pipeline:
     operators:
diff --git a/docs/sphinx_doc/source/tutorial/trinity_programming_guide.md b/docs/sphinx_doc/source/tutorial/trinity_programming_guide.md
index 516b802084..c87ec8779f 100644
--- a/docs/sphinx_doc/source/tutorial/trinity_programming_guide.md
+++ b/docs/sphinx_doc/source/tutorial/trinity_programming_guide.md
@@ -83,7 +83,7 @@ buffer:
   explorer_input:
     taskset:
       default_workflow: "math_workflow"
-      path: "/PATH/TO/FILE/DIR"
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       format:
         prompt_key: "question"
         response_key: "answer"
diff --git a/examples/RAFT_alfworld/RAFT_alfworld_7B.yaml b/examples/RAFT_alfworld/RAFT_alfworld_7B.yaml
index 78d5793d4f..3dbf42645f 100644
--- a/examples/RAFT_alfworld/RAFT_alfworld_7B.yaml
+++ b/examples/RAFT_alfworld/RAFT_alfworld_7B.yaml
@@ -1,12 +1,12 @@
 project: "Trinity-RFT-RAFT-alfworld"
 name: "qwen2.5-7B-RAFT-alfworld"
 mode: both
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/RAFT_ALFWORLD/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: raft
   repeat_times: 1
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 4096
   max_model_len: 20480
 cluster:
@@ -19,7 +19,7 @@ buffer:
     taskset:
       name: alfworld-train
       storage_type: file
-      path: '/PATH/TO/ALFWORLD_DATA/'
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       split: 'train'
       format:
         prompt_key: 'game_file'
@@ -31,7 +31,7 @@ buffer:
     eval_tasksets:
       - name: alfworld-eval
         storage_type: file
-        path: '/PATH/TO/ALFWORLD_DATA/'
+        path: ${oc.env:TRINITY_TASKSET_PATH}
         split: 'test'
         format:
           prompt_key: 'game_file'
diff --git a/examples/RAFT_alfworld/RAFT_reflect_alfworld_7B.yaml b/examples/RAFT_alfworld/RAFT_reflect_alfworld_7B.yaml
index fc721b68f7..ed32dab857 100644
--- a/examples/RAFT_alfworld/RAFT_reflect_alfworld_7B.yaml
+++ b/examples/RAFT_alfworld/RAFT_reflect_alfworld_7B.yaml
@@ -1,12 +1,12 @@
 project: "Trinity-RFT-RAFT-reflect-alfworld"
 name: "qwen2.5-7B-RAFT-reflect-alfworld"
 mode: both
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/RAFT_REFLECT_ALFWORLD/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: raft
   repeat_times: 1
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 4096
   max_model_len: 20480
 cluster:
@@ -19,7 +19,7 @@ buffer:
     taskset:
       name: alfworld-train
       storage_type: file
-      path: '/PATH/TO/ALFWORLD_DATA/'
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       split: 'train'
       format:
         prompt_key: 'game_file'
@@ -31,7 +31,7 @@ buffer:
     eval_tasksets:
       - name: alfworld-eval
         storage_type: file
-        path: '/PATH/TO/ALFWORLD_DATA/'
+        path: ${oc.env:TRINITY_TASKSET_PATH}
         split: 'test'
         format:
           prompt_key: 'game_file'
diff --git a/examples/RAFT_alfworld/README.md b/examples/RAFT_alfworld/README.md
index 291a150005..1f0835fdb9 100644
--- a/examples/RAFT_alfworld/README.md
+++ b/examples/RAFT_alfworld/README.md
@@ -22,7 +22,7 @@ python examples/grpo_alfworld/get_alfworld_data.py
 
 ### Configuration
 Before running, make sure to update the following paths in the YAML files:
-- `model.model_path`: Replace with your model path (e.g., `/PATH/TO/MODEL/`)
+- `model.model_path`: Replace with your model path (e.g., `Qwen/Qwen2.5-7B-Instruct`)
 - `buffer.explorer_input.taskset.path`: Replace with your alfworld dataset path
 - `buffer.explorer_input.eval_tasksets[0].path`: Replace with your alfworld dataset path
 - `checkpoint_root_dir`: Replace with your desired checkpoint directory
diff --git a/examples/agentscope_tool_react/agentscope_tool_react_dapo.yaml b/examples/agentscope_tool_react/agentscope_tool_react_dapo.yaml
index 0ff1792fce..cfc6e80152 100644
--- a/examples/agentscope_tool_react/agentscope_tool_react_dapo.yaml
+++ b/examples/agentscope_tool_react/agentscope_tool_react_dapo.yaml
@@ -1,12 +1,12 @@
 project: "Trinity-RFT-dapo-reactv2"
 name: "Qwen3-8B-dapo-reactv2"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
   advantage_fn: step_wise_grpo
 model:
-  model_path: /PATH/TO/MODEL/Qwen3-8B
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen3-8B}
   max_response_tokens: 16384
   max_model_len: 24576
 cluster:
diff --git a/examples/agentscope_tool_react/agentscope_tool_react_gsm8k.yaml b/examples/agentscope_tool_react/agentscope_tool_react_gsm8k.yaml
index 07fe2a2756..a40e96fd8e 100644
--- a/examples/agentscope_tool_react/agentscope_tool_react_gsm8k.yaml
+++ b/examples/agentscope_tool_react/agentscope_tool_react_gsm8k.yaml
@@ -1,12 +1,12 @@
 project: "Trinity-RFT-gsm8k-reactv2"
 name: "Qwen3-4B-gsm8k-reactv2"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
   advantage_fn: step_wise_grpo
 model:
-  model_path: /PATH/TO/MODEL/Qwen3-4B
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen3-4B}
   max_response_tokens: 16384
   max_model_len: 24576
 cluster:
diff --git a/examples/asymre_gsm8k/gsm8k.yaml b/examples/asymre_gsm8k/gsm8k.yaml
index cde75174e8..7977a0957c 100644
--- a/examples/asymre_gsm8k/gsm8k.yaml
+++ b/examples/asymre_gsm8k/gsm8k.yaml
@@ -4,9 +4,9 @@
 
 project: "Trinity-RFT-GSM8K"
 name: asymre_gsm8k
-checkpoint_root_dir:  /PATH/TO/CHECKPOINT/
+checkpoint_root_dir:  ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 algorithm:
@@ -26,7 +26,8 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
       split: train
       format:
         prompt_key: question
@@ -36,7 +37,8 @@ buffer:
     eval_tasksets:
       - name: gsm8k-eval
         storage_type: file
-        path: /PATH/TO/DATASET/
+        path: 'openai/gsm8k'
+        subset_name: 'main'
         split: test
         format:
           prompt_key: question
diff --git a/examples/asymre_math/math.yaml b/examples/asymre_math/math.yaml
index 888ef61bf0..b7ca4f5e61 100644
--- a/examples/asymre_math/math.yaml
+++ b/examples/asymre_math/math.yaml
@@ -3,9 +3,9 @@
 # https://arxiv.org/abs/2506.20520.
 project: "Trinity-RFT-MATH"
 name: asymre_math
-checkpoint_root_dir:  /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL/ # the path to your model
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-3B-Instruct} # the path to your model
   max_response_tokens: 1024
   max_model_len: 1280
 algorithm:
@@ -26,7 +26,7 @@ buffer:
     taskset:
       name: math
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       format:
         prompt_key: 'problem'
         response_key: 'solution'
@@ -37,7 +37,7 @@ buffer:
     eval_tasksets:
       - name: math
         storage_type: file
-        path: /PATH/TO/DATASET/
+        path: ${oc.env:TRINITY_TASKSET_PATH}
         split: 'test'
         format:
           prompt_key: 'problem'
diff --git a/examples/async_gsm8k/explorer.yaml b/examples/async_gsm8k/explorer.yaml
index 45e2f74f6a..07c21ef043 100644
--- a/examples/async_gsm8k/explorer.yaml
+++ b/examples/async_gsm8k/explorer.yaml
@@ -1,12 +1,12 @@
 project: "Trinity-RFT-gsm8k"
 name: "async-qwen2.5-1.5B-gsm8k"
 mode: explore
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
@@ -19,7 +19,8 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
       split: train
       format:
         prompt_key: 'question'
diff --git a/examples/async_gsm8k/trainer.yaml b/examples/async_gsm8k/trainer.yaml
index 90e4663f98..ae60168ad7 100644
--- a/examples/async_gsm8k/trainer.yaml
+++ b/examples/async_gsm8k/trainer.yaml
@@ -1,12 +1,12 @@
 project: "Trinity-RFT-gsm8k"
 name: "async-qwen2.5-1.5B-gsm8k"
 mode: train
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
@@ -19,7 +19,9 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
+      split: 'train'
       format:
         prompt_key: 'question'
         response_key: 'answer'
diff --git a/examples/cispo_gsm8k/gsm8k.yaml b/examples/cispo_gsm8k/gsm8k.yaml
index 8e01e321c5..02d787ef9b 100644
--- a/examples/cispo_gsm8k/gsm8k.yaml
+++ b/examples/cispo_gsm8k/gsm8k.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-gsm8k"
 name: "qwen2.5-1.5B-gsm8k-cispo"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: cispo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
@@ -45,7 +45,7 @@ buffer:
     # sft_warmup_dataset: # Uncomment these to enable sft warmup
     #   name: warmup_data
     #   storage_type: file
-    #   path: '/PATH/TO/WARMUP_DATA/'
+    #   path: ${oc.env:TRINITY_SFT_DATASET_PATH}
 explorer:
   eval_interval: 50
   runner_per_model: 8
diff --git a/examples/dapo_math/dapo.yaml b/examples/dapo_math/dapo.yaml
index 333ccf9c03..1bd7b1d7ed 100644
--- a/examples/dapo_math/dapo.yaml
+++ b/examples/dapo_math/dapo.yaml
@@ -1,8 +1,8 @@
 project: Trinity-RFT-example
 name: dapo
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-32B-Instruct}
   max_response_tokens: 20480
   max_model_len: 21504
 algorithm:
@@ -40,7 +40,7 @@ buffer:
     eval_tasksets:
     - name: AIME2024
       storage_type: file
-      path: /PATH/TO/AIME2024/
+      path: ${oc.env:TRINITY_TASKSET_PATH}  # e.g. path to AIME2024
       split: 'test'
       repeat_times: 32
       format:
diff --git a/examples/dpo_human_in_the_loop/dpo.yaml b/examples/dpo_human_in_the_loop/dpo.yaml
index bd3ab3a6cb..b202c5e211 100644
--- a/examples/dpo_human_in_the_loop/dpo.yaml
+++ b/examples/dpo_human_in_the_loop/dpo.yaml
@@ -41,9 +41,9 @@ algorithm:
   kl_loss_fn: k1
   kl_loss_fn_args:
     kl_coef: 0.1
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1536
 cluster:
diff --git a/examples/dpo_humanlike/dpo.yaml b/examples/dpo_humanlike/dpo.yaml
index 81042ff431..c932fa0a43 100644
--- a/examples/dpo_humanlike/dpo.yaml
+++ b/examples/dpo_humanlike/dpo.yaml
@@ -6,9 +6,9 @@ algorithm:
   kl_loss_fn: k1
   kl_loss_fn_args:
     kl_coef: 0.1
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1536
 cluster:
@@ -22,7 +22,7 @@ buffer:
       name: dpo_buffer
       storage_type: file
       enable_progress_bar: True
-      path: /PATH/TO/DATASET/
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       format:
         prompt_type: plaintext # plaintext/messages
         prompt_key: prompt
diff --git a/examples/grpo_alfworld/alfworld.yaml b/examples/grpo_alfworld/alfworld.yaml
index ca19267ad6..ed993790d4 100644
--- a/examples/grpo_alfworld/alfworld.yaml
+++ b/examples/grpo_alfworld/alfworld.yaml
@@ -1,11 +1,11 @@
 project: "ALFWORLD"
 name: "ALFWORLD_RFT"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/ALFWORLD_RFT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 16384
   max_model_len: 20480
 cluster:
@@ -43,7 +43,7 @@ explorer:
     gpu_memory_utilization: 0.7
     enable_chunked_prefill: true
   env_vars:
-    TMPDIR: /PATH/TO/ALFWORLD_TMP_DIR
+    TMPDIR: ${oc.env:TMPDIR,/tmp}
 synchronizer:
   sync_method: 'nccl'
   sync_interval: 5
diff --git a/examples/grpo_alfworld_general_multi_step/alfworld.yaml b/examples/grpo_alfworld_general_multi_step/alfworld.yaml
index 8829b3b77f..f93f028420 100644
--- a/examples/grpo_alfworld_general_multi_step/alfworld.yaml
+++ b/examples/grpo_alfworld_general_multi_step/alfworld.yaml
@@ -1,12 +1,12 @@
 project: "ALFWORLD"
 name: "Step_Wise_Alfworld"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/ALFWORLD_RFT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 16
   advantage_fn: step_wise_grpo
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 16384
   max_model_len: 20480
 cluster:
@@ -50,7 +50,7 @@ explorer:
     gpu_memory_utilization: 0.7
     enable_chunked_prefill: true
   env_vars:
-    TMPDIR: /PATH/TO/ALFWORLD_TMP_DIR
+    TMPDIR: ${oc.env:TMPDIR,/tmp}
 synchronizer:
   sync_style: dynamic_by_explorer
   sync_method: 'nccl'
diff --git a/examples/grpo_email_search/email_search.yaml b/examples/grpo_email_search/email_search.yaml
index 6b861ae96c..67f0cb098e 100644
--- a/examples/grpo_email_search/email_search.yaml
+++ b/examples/grpo_email_search/email_search.yaml
@@ -1,12 +1,12 @@
 project: "Trinity_Multi_Step"
 name: "Email_Example"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
   advantage_fn: grpo
 model:
-  model_path: /PATH/TO/Qwen3-4B-Instruct-2507
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen3-4B-Instruct-2507}
   max_response_tokens: 4096
   max_model_len: 20480
 cluster:
@@ -20,7 +20,7 @@ buffer:
     taskset:
       name: enron_train
       storage_type: file
-      path: '/PATH/TO/ENRON_EMAIL_QA_PAIRS/'
+      path: ${oc.env:TRINITY_TASKSET_PATH}  # e.g. path to ENRON_EMAIL_QA_PAIRS
       split: train
       format:
         prompt_key: 'question'
@@ -35,7 +35,7 @@ buffer:
     eval_tasksets:
     - name: enron_test
       storage_type: file
-      path: '/PATH/TO/ENRON_EMAIL_QA_PAIRS/'
+      path: ${oc.env:TRINITY_TASKSET_PATH}  # e.g. path to ENRON_EMAIL_QA_PAIRS
       split: test
       format:
         prompt_key: 'question'
@@ -73,7 +73,7 @@ explorer:
     gpu_memory_utilization: 0.7
     enable_chunked_prefill: true
   auxiliary_models:
-    - model_path: /PATH/TO/Qwen3-30B-A3B-Instruct-2507
+    - model_path: Qwen/Qwen3-30B-A3B-Instruct-2507
       engine_num: 1
       tensor_parallel_size: 2
       enable_thinking: false
diff --git a/examples/grpo_gsm8k/gsm8k.yaml b/examples/grpo_gsm8k/gsm8k.yaml
index 327d6fb10b..78d9a2d85b 100644
--- a/examples/grpo_gsm8k/gsm8k.yaml
+++ b/examples/grpo_gsm8k/gsm8k.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-gsm8k"
 name: "qwen2.5-1.5B-gsm8k"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
@@ -45,7 +45,7 @@ buffer:
     # sft_warmup_dataset: # Uncomment these to enable sft warmup
     #   name: warmup_data
     #   storage_type: file
-    #   path: '/PATH/TO/WARMUP_DATA/'
+    #   path: ${oc.env:TRINITY_SFT_DATASET_PATH}
 explorer:
   eval_interval: 50
   runner_per_model: 8
diff --git a/examples/grpo_gsm8k_experience_pipeline/gsm8k.yaml b/examples/grpo_gsm8k_experience_pipeline/gsm8k.yaml
index 18aa511ac8..e78f7abd9f 100644
--- a/examples/grpo_gsm8k_experience_pipeline/gsm8k.yaml
+++ b/examples/grpo_gsm8k_experience_pipeline/gsm8k.yaml
@@ -1,6 +1,6 @@
 project: "Trinity-RFT-gsm8k-exp-pipe"
 name: "qwen2.5-1.5B-gsm8k-exp-pipe"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 
 service:
   data_juicer:
@@ -25,7 +25,7 @@ algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
diff --git a/examples/grpo_gsm8k_ruler/gsm8k_ruler.yaml b/examples/grpo_gsm8k_ruler/gsm8k_ruler.yaml
index f578f4df67..2adddd0b6e 100644
--- a/examples/grpo_gsm8k_ruler/gsm8k_ruler.yaml
+++ b/examples/grpo_gsm8k_ruler/gsm8k_ruler.yaml
@@ -1,13 +1,13 @@
 project: "Trinity-RFT-gsm8k-ruler"
 name: "qwen2.5-1.5B-gsm8k-ruler"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   advantage_fn_args:
     std_threshold: 0.0001  # effectively zero
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
@@ -53,7 +53,7 @@ explorer:
     dtype: bfloat16
     seed: 42
   auxiliary_models:
-    - model_path: /PATH/TO/Qwen2.5-32B-Instruct
+    - model_path: Qwen/Qwen2.5-32B-Instruct
       engine_num: 1
       tensor_parallel_size: 2
       enable_thinking: false
diff --git a/examples/grpo_gsm8k_task_pipeline/gsm8k.yaml b/examples/grpo_gsm8k_task_pipeline/gsm8k.yaml
index 0dd9397efb..bf7d996feb 100644
--- a/examples/grpo_gsm8k_task_pipeline/gsm8k.yaml
+++ b/examples/grpo_gsm8k_task_pipeline/gsm8k.yaml
@@ -1,6 +1,6 @@
 project: "Trinity-RFT-gsm8k-task-pipeline"
 name: "qwen2.5-1.5B-gsm8k-task-pipeline"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
@@ -20,11 +20,11 @@ data_processor:
           input_keys: ["question", "answer"]
           field_names: ["Question", "Answer"]
     inputs:  # the output will be set to the explorer input automatically
-      - /PATH/TO/GSM8K/DATA/FILE
+      - ${oc.env:TRINITY_TASKSET_PATH}
     target_fields: ["question", "answer"]
 
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
diff --git a/examples/grpo_math/math.yaml b/examples/grpo_math/math.yaml
index ead393ffc0..601e3dea29 100644
--- a/examples/grpo_math/math.yaml
+++ b/examples/grpo_math/math.yaml
@@ -1,8 +1,8 @@
 project: grpo_math
 name: grpo_math_example
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 3072
   max_model_len: 4096
 algorithm:
@@ -18,7 +18,7 @@ buffer:
     taskset:
       name: math
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       format:
         prompt_key: 'question'
         response_key: 'gt_answer'
diff --git a/examples/grpo_sciworld/sciworld.yaml b/examples/grpo_sciworld/sciworld.yaml
index 8d7bc08ecd..8b2c3703d3 100644
--- a/examples/grpo_sciworld/sciworld.yaml
+++ b/examples/grpo_sciworld/sciworld.yaml
@@ -1,11 +1,11 @@
 project: "sciworld"
 name: "sciworld_RFT"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 16384
   max_model_len: 20480
 cluster:
diff --git a/examples/grpo_toolcall/toolace.yaml b/examples/grpo_toolcall/toolace.yaml
index e6d324387f..1bce1884c4 100644
--- a/examples/grpo_toolcall/toolace.yaml
+++ b/examples/grpo_toolcall/toolace.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-toolace"
 name: "qwen2.5-7B-toolace"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen1.5-7B-Chat}
   max_response_tokens: 8192
   max_model_len: 12288
 cluster:
diff --git a/examples/grpo_vlm/vlm.yaml b/examples/grpo_vlm/vlm.yaml
index 7ce58841a1..890fed3349 100644
--- a/examples/grpo_vlm/vlm.yaml
+++ b/examples/grpo_vlm/vlm.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT"
 name: "grpo_vlm"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/Qwen2.5-VL-3B-Instruct
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-VL-3B-Instruct}
   max_response_tokens: 1024
   max_model_len: 20480
 cluster:
diff --git a/examples/grpo_webshop/webshop.yaml b/examples/grpo_webshop/webshop.yaml
index 4ce466aa29..7b7b7f0611 100644
--- a/examples/grpo_webshop/webshop.yaml
+++ b/examples/grpo_webshop/webshop.yaml
@@ -1,11 +1,11 @@
 project: "WEBSHOP"
 name: "WEBSHOP_RFT"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: grpo
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 16384
   max_model_len: 20480
 cluster:
diff --git a/examples/mix_chord/mix_chord.yaml b/examples/mix_chord/mix_chord.yaml
index 46d0b8caff..aaa3f6eeee 100644
--- a/examples/mix_chord/mix_chord.yaml
+++ b/examples/mix_chord/mix_chord.yaml
@@ -1,6 +1,6 @@
 project: "mix_chord"
 name: "test_mix_chord"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: mix_chord
   repeat_times: 8 # or 16 for better performance in math related tasks
@@ -23,7 +23,7 @@ algorithm:
     train_batch_size_expert: 64
     train_batch_size_usual: 256 # 32 batchsize * 8 repeat times
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 10240
   max_model_len: 11264
 cluster:
@@ -37,7 +37,7 @@ buffer:
     taskset:
       name: openr1_data_filtered_int
       storage_type: file
-      path: /PATH/TO/RL_DATASET
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       format:
         prompt_key: 'problem'
         response_key: 'answer'
@@ -57,7 +57,7 @@ buffer:
       total_epochs: 25
       name: SFT_data
       storage_type: file
-      path: /PATH/TO/SFT_DATASET
+      path: ${oc.env:TRINITY_SFT_DATASET_PATH,open-r1/Mixture-of-Thoughts}
       split: 'train'
       format:
         prompt_type: messages
diff --git a/examples/mix_math/mix_math.yaml b/examples/mix_math/mix_math.yaml
index 043c57d55b..68d45d9696 100644
--- a/examples/mix_math/mix_math.yaml
+++ b/examples/mix_math/mix_math.yaml
@@ -1,6 +1,6 @@
 project: "mix_math"
 name: "expert0.20_mu0.1"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: mix
   repeat_times: 8
@@ -17,7 +17,7 @@ algorithm:
     train_batch_size_expert: 64
     train_batch_size_usual: 256
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 10240
   max_model_len: 11264
 cluster:
@@ -31,7 +31,7 @@ buffer:
     taskset:
       name: math_train
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       split: 'train'
       format:
         prompt_key: 'problem'
@@ -42,7 +42,7 @@ buffer:
     eval_tasksets:
     - name: math_eval
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: ${oc.env:TRINITY_TASKSET_PATH}
       split: 'test'
       format:
         prompt_key: 'problem'
@@ -52,12 +52,12 @@ buffer:
     experience_buffer:
       name: math_buffer
       storage_type: queue
-      path: /PATH/TO/BUFFER/
+      path: ${oc.env:TRINITY_BUFFER_PATH,./exp_buffer}
     sft_warmup_dataset:
       total_epochs: 10
       name: math_sft
       storage_type: file
-      path: /PATH/TO/EXPERT_DATA/
+      path: ${oc.env:TRINITY_SFT_DATASET_PATH,open-r1/Mixture-of-Thoughts}
       split: 'train'
       format:
         prompt_type: messages
diff --git a/examples/opmd_gsm8k/opmd_gsm8k.yaml b/examples/opmd_gsm8k/opmd_gsm8k.yaml
index 81d1ae50c9..cdce17f58d 100644
--- a/examples/opmd_gsm8k/opmd_gsm8k.yaml
+++ b/examples/opmd_gsm8k/opmd_gsm8k.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-gsm8k-test-opmd"
 name: "opmd_test"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: opmd
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 512
   max_model_len: 1024
 cluster:
diff --git a/examples/ppo_countdown/countdown.yaml b/examples/ppo_countdown/countdown.yaml
index 7ce9a604d4..9df019ca61 100644
--- a/examples/ppo_countdown/countdown.yaml
+++ b/examples/ppo_countdown/countdown.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-countdown"
 name: "qwen2.5-1.5B-countdown"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: ppo
   repeat_times: 5
 model:
-  model_path: '/PATH/TO/MODEL/CHECKPOINT/'
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
diff --git a/examples/ppo_countdown_megatron/countdown.yaml b/examples/ppo_countdown_megatron/countdown.yaml
index d0996bcedd..f0aa2724b2 100644
--- a/examples/ppo_countdown_megatron/countdown.yaml
+++ b/examples/ppo_countdown_megatron/countdown.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-countdown"
 name: "qwen2.5-1.5B-countdown"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: ppo
   repeat_times: 5
 model:
-  model_path: '/PATH/TO/MODEL/CHECKPOINT/'
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
diff --git a/examples/sft_mot/sft.yaml b/examples/sft_mot/sft.yaml
index 88d78cb06f..4542310e56 100644
--- a/examples/sft_mot/sft.yaml
+++ b/examples/sft_mot/sft.yaml
@@ -1,11 +1,11 @@
 mode: train
 project: "Trinity-RFT-example"
 name: "sft_mot"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: sft
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-7B-Instruct}
   max_response_tokens: 10240
   max_model_len: 10752
 cluster:
diff --git a/examples/sppo_gsm8k/gsm8k.yaml b/examples/sppo_gsm8k/gsm8k.yaml
index 49c886fde4..700f89029d 100644
--- a/examples/sppo_gsm8k/gsm8k.yaml
+++ b/examples/sppo_gsm8k/gsm8k.yaml
@@ -4,9 +4,9 @@
 
 project: "Trinity-RFT-GSM8K"
 name: sppo_gsm8k
-checkpoint_root_dir:  /PATH/TO/CHECKPOINT/
+checkpoint_root_dir:  ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 algorithm:
@@ -24,7 +24,8 @@ buffer:
     taskset:
       name: gsm8k
       storage_type: file
-      path: /PATH/TO/DATASET/
+      path: 'openai/gsm8k'
+      subset_name: 'main'
       split: train
       format:
         prompt_key: question
@@ -34,7 +35,8 @@ buffer:
     eval_tasksets:
       - name: gsm8k-eval
         storage_type: file
-        path: /PATH/TO/DATASET/
+        path: 'openai/gsm8k'
+        subset_name: 'main'
         split: test
         format:
           prompt_key: question
diff --git a/examples/topr_gsm8k/gsm8k.yaml b/examples/topr_gsm8k/gsm8k.yaml
index 98aec1e7d7..bb7cc094e2 100644
--- a/examples/topr_gsm8k/gsm8k.yaml
+++ b/examples/topr_gsm8k/gsm8k.yaml
@@ -1,11 +1,11 @@
 project: "Trinity-RFT-gsm8k"
 name: "qwen2.5-1.5B-gsm8k-topr"
-checkpoint_root_dir: /PATH/TO/CHECKPOINT/
+checkpoint_root_dir: ${oc.env:TRINITY_CHECKPOINT_ROOT_DIR,./checkpoints}
 algorithm:
   algorithm_type: topr
   repeat_times: 8
 model:
-  model_path: /PATH/TO/MODEL/
+  model_path: ${oc.env:TRINITY_MODEL_PATH,Qwen/Qwen2.5-1.5B-Instruct}
   max_response_tokens: 1024
   max_model_len: 1280
 cluster:
diff --git a/tests/common/vllm_test.py b/tests/common/vllm_test.py
index c442f60f6e..d9c54deda5 100644
--- a/tests/common/vllm_test.py
+++ b/tests/common/vllm_test.py
@@ -1,11 +1,16 @@
-import os
 import unittest
 
 import torch
 from parameterized import parameterized_class
 from transformers import AutoTokenizer
 
-from tests.tools import RayUnittestBase, RayUnittestBaseAysnc, get_template_config
+from tests.tools import (
+    RayUnittestBase,
+    RayUnittestBaseAysnc,
+    get_api_model_path,
+    get_model_path,
+    get_template_config,
+)
 from trinity.common.models import create_inference_models
 from trinity.common.models.model import ModelWrapper
 from trinity.common.models.utils import (
@@ -13,25 +18,6 @@
     tokenize_and_mask_messages_hf,
 )
 
-
-def get_model_path() -> str:
-    path = os.environ.get("MODEL_PATH")
-    if not path:
-        raise EnvironmentError(
-            "Please set `export MODEL_PATH=<your_model_checkpoint_dir>` before running this test."
-        )
-    return path
-
-
-def get_api_model_path() -> str:
-    path = os.environ.get("API_MODEL_PATH")
-    if not path:
-        raise EnvironmentError(
-            "Please set `export API_MODEL_PATH=<your_api_model_checkpoint_dir>` before running this test."
-        )
-    return path
-
-
 DEBUG = False
 
 
diff --git a/tests/tools.py b/tests/tools.py
index 02c3cd5f2c..7dc94bf881 100644
--- a/tests/tools.py
+++ b/tests/tools.py
@@ -7,7 +7,15 @@
 from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
 
 from trinity.common.config import Config, FormatConfig, StorageConfig, load_config
-from trinity.common.constants import PromptType
+from trinity.common.constants import (
+    CHECKPOINT_ROOT_DIR_ENV_VAR,
+    MODEL_PATH_ENV_VAR,
+    PromptType,
+)
+
+API_MODEL_PATH_ENV_VAR = "TRINITY_API_MODEL_PATH"
+VLM_MODEL_PATH_ENV_VAR = "TRINITY_VLM_MODEL_PATH"
+SFT_DATASET_PATH_ENV_VAR = "TRINITY_SFT_DATASET_PATH"
 
 
 def get_template_config() -> Config:
@@ -21,28 +29,37 @@ def get_template_config() -> Config:
 
 
 def get_model_path() -> str:
-    path = os.environ.get("MODEL_PATH")
+    path = os.environ.get(MODEL_PATH_ENV_VAR)
+    if not path:
+        raise EnvironmentError(
+            f"Please set `export {MODEL_PATH_ENV_VAR}=<your_model_dir>` before running this test."
+        )
+    return path
+
+
+def get_api_model_path() -> str:
+    path = os.environ.get(API_MODEL_PATH_ENV_VAR)
     if not path:
         raise EnvironmentError(
-            "Please set `export MODEL_PATH=<your_model_dir>` before running this test."
+            f"Please set `export {API_MODEL_PATH_ENV_VAR}=<your_api_model_checkpoint_dir>` before running this test."
         )
     return path
 
 
 def get_checkpoint_path() -> str:
-    path = os.environ.get("CHECKPOINT_PATH")
+    path = os.environ.get(CHECKPOINT_ROOT_DIR_ENV_VAR)
     if not path:
         raise EnvironmentError(
-            "Please set `export CHECKPOINT_PATH=<your_checkpoint_dir>` before running this test."
+            f"Please set `export {CHECKPOINT_ROOT_DIR_ENV_VAR}=<your_checkpoint_dir>` before running this test."
         )
     return path
 
 
 def get_vision_languge_model_path() -> str:
-    path = os.environ.get("VLM_MODEL_PATH")
+    path = os.environ.get(VLM_MODEL_PATH_ENV_VAR)
     if not path:
         raise EnvironmentError(
-            "Please set `export VLM_MODEL_PATH=<your_model_dir>` before running this test."
+            f"Please set `export {VLM_MODEL_PATH_ENV_VAR}=<your_model_dir>` before running this test."
         )
     return path
 
diff --git a/trinity/common/constants.py b/trinity/common/constants.py
index aa35472c5b..10fff98d99 100644
--- a/trinity/common/constants.py
+++ b/trinity/common/constants.py
@@ -10,6 +10,10 @@
 ROLLOUT_WEIGHT_SYNC_GROUP_NAME = "rollout_weight_sync"
 
 # trinity env var names
+CHECKPOINT_ROOT_DIR_ENV_VAR = "TRINITY_CHECKPOINT_ROOT_DIR"
+MODEL_PATH_ENV_VAR = "TRINITY_MODEL_PATH"
+TASKSET_PATH_ENV_VAR = "TRINITY_TASKSET_PATH"
+BUFFER_PATH_ENV_VAR = "TRINITY_BUFFER_PATH"
 PLUGIN_DIRS_ENV_VAR = "TRINITY_PLUGIN_DIRS"
 LOG_DIR_ENV_VAR = "TRINITY_LOG_DIR"  # log dir
 LOG_LEVEL_ENV_VAR = "TRINITY_LOG_LEVEL"  # global log level