Adapt the supplied project README into the model card and update repository links

.gitattributes

@@ -266,3 +266,12 @@ MATH/SLS/Qwen3-8B/checkpoint-1475/tokenizer.json filter=lfs diff=lfs merge=lfs -
 MATH/SLS/Qwen3-8B/checkpoint-1750/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 MATH/SLS/Qwen3-8B/checkpoint-1000/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 MATH/SLS/Qwen3-8B/checkpoint-1600/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+assets/figures/previews/Figure1-1.png filter=lfs diff=lfs merge=lfs -text
+assets/figures/previews/Figure2-1.png filter=lfs diff=lfs merge=lfs -text
+assets/figures/previews/Figure3-1.png filter=lfs diff=lfs merge=lfs -text
+assets/figures/previews/Figure4-1.png filter=lfs diff=lfs merge=lfs -text
+assets/figures/previews/Figure5-1.png filter=lfs diff=lfs merge=lfs -text
+assets/figures/Figure1.pdf filter=lfs diff=lfs merge=lfs -text
+assets/figures/Figure2.pdf filter=lfs diff=lfs merge=lfs -text
+assets/figures/Figure4.pdf filter=lfs diff=lfs merge=lfs -text
+assets/figures/Figure5.pdf filter=lfs diff=lfs merge=lfs -text

README.md

@@ -15,42 +15,159 @@ tags:
 - qwen3
 ---
 
-# JCA: main-experiment checkpoints
+<h1 align="center">Reasoning across Sliced Models via Collaborative Learning</h1>
+
+<p align="center">
+  <a href="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/Figure1.pdf"><img src="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/previews/Figure1-1.png" alt="Collaborative Learning overview: model slicing, protocol distillation, and pyramid supervision" width="960"></a>
+</p>
+
+<p align="center">
+  <a href="#-overview">Overview</a>
+  &nbsp;·&nbsp;
+  <a href="#-evaluation">Evaluation</a>
+  &nbsp;·&nbsp;
+  <a href="#-workflow">Workflow</a>
+  &nbsp;·&nbsp;
+  <a href="#-repository-layout">Repository layout</a>
+  &nbsp;·&nbsp;
+  <a href="#-quickstart">Quickstart</a>
+  &nbsp;·&nbsp;
+  <a href="#-code-and-experiment-entry-points">Code and experiment entry points</a>
+</p>
+
+## ✨ Overview
+
+**Collaborative Learning (CL)** studies how sliced models can collaborate on reasoning while keeping the overall parameter budget comparable to a single larger model. The method combines model slicing, protocol distillation, and pyramid supervision so that models can build on shared task state across multiple reasoning steps.
+
+This model repository releases the **rank-64 LoRA adapters for the heterogeneous CL main experiments**, using Qwen3-1.7B, Qwen3-4B, and Qwen3-8B. It covers four task families—multi-hop question answering, mathematics, code generation, and constrained generation—across MuSiQue, GSM-Hard, MATH, MultiPL-E, and Conifer. The [companion code repository](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c) contains the implementation, benchmark scorers, result records, configurations, and analysis artifacts.
+
+<p align="center">
+  <a href="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/Figure2.pdf">
+    <img src="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/previews/Figure2-1.png" alt="Model slicing, protocol distillation, and pyramid supervision" width="920">
+  </a>
+</p>
+
+## 🧪 Evaluation
+
+We evaluate CL on four task families and five benchmarks:
+
+| Task family | Benchmark | What it measures | Checkpoints |
+| --- | --- | --- | --- |
+| Multi-hop QA | [MuSiQue](https://github.com/StonyBrookNLP/musique) | Compositional retrieval and reasoning | [MuSiQue](https://huggingface.co/SJTUFreshman/collaborativelearning/tree/main/MuSiQue) |
+| Mathematical reasoning | **GSM-Hard** | Robust numerical reasoning | [GSM-Hard](https://huggingface.co/SJTUFreshman/collaborativelearning/tree/main/GSM-Hard) |
+| Mathematical reasoning | **MATH** | Competition-level problem solving | [MATH](https://huggingface.co/SJTUFreshman/collaborativelearning/tree/main/MATH) |
+| Code generation | [MultiPL-E](https://github.com/nuprl/MultiPL-E) | Pass@1 across eight languages | [MultiPL-E](https://huggingface.co/SJTUFreshman/collaborativelearning/tree/main/MultiPL-E) |
+| Constrained generation | **Conifer** | Coverage and explicit constraint satisfaction | [Conifer](https://huggingface.co/SJTUFreshman/collaborativelearning/tree/main/Conifer) |
+
+### 📊 Results
+
+The table below compares CL with the main single-model and multi-model baselines. All values are percentages reported in the paper, not a new evaluation of this weight release. MuSiQue, GSM-Hard, and MATH use **accuracy / F1**; MultiPL-E uses **weighted / macro-language pass@1**; Conifer uses **Coverage / Explicit**.
+
+| Method | MuSiQue | GSM-Hard | MATH | MultiPL-E | Conifer |
+| --- | ---: | ---: | ---: | ---: | ---: |
+| Qwen3-14B | 39.55 / 49.70 | 68.94 / 80.69 | 74.60 / 82.86 | 55.18 / 57.34 | 88.00 / 95.72 |
+| SE-RL | 41.76 / 52.73 | 68.18 / 79.67 | 76.00 / 80.70 | 60.43 / 62.55 | 88.00 / 95.81 |
+| MAD | 39.51 / 50.59 | 66.67 / 74.18 | 72.40 / 81.32 | 55.03 / 55.35 | 88.55 / 95.84 |
+| AgentVerse | 41.99 / 52.85 | 71.97 / 78.44 | 68.20 / 79.42 | 59.25 / 59.83 | 87.74 / 95.86 |
+| AFlow | 41.83 / 50.36 | 68.18 / 76.04 | 65.00 / 76.53 | 42.83 / 42.46 | 87.92 / 96.51 |
+| GPTSwarm | 36.20 / 47.37 | 70.45 / 78.39 | 69.20 / 79.90 | 51.63 / 52.43 | 88.17 / 95.96 |
+| MAGRPO | 40.13 / 51.74 | 65.91 / 76.44 | 65.00 / 76.01 | 49.93 / 49.93 | 88.58 / 95.82 |
+| MAPoRL | 40.59 / 52.74 | 63.64 / 75.34 | 70.80 / 80.87 | 53.25 / 53.27 | 86.37 / 89.19 |
+| AT-GRPO | 41.54 / 53.51 | 64.39 / 76.86 | 74.20 / 83.70 | 57.91 / 58.74 | 86.60 / 95.85 |
+| Homo. MAS | 26.40 / 32.82 | 71.21 / 79.93 | 68.20 / 76.41 | 56.36 / 57.20 | 88.02 / 94.27 |
+| Homo. CL | 36.00 / 47.19 | 60.61 / 76.51 | 68.60 / 77.95 | 53.33 / 52.65 | 88.15 / 95.22 |
+| **CL (ours)** | **42.37 / 53.43** | **71.97 / 81.15** | **78.40 / 85.25** | **59.39 / 59.63** | **88.90 / 95.90** |
+
+<p align="center">
+  <a href="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/Figure3.pdf">
+    <img src="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/previews/Figure3-1.png" alt="Ablation study across the five benchmarks" width="760">
+  </a>
+</p>
+
+The evaluation records contain 2,417 MuSiQue examples, 132 GSM-Hard examples, 500 MATH problems, 1,352 MultiPL-E test problems, and 1,402 Conifer trajectories.
+
+CL is best or joint-best on **6 of 10 metrics**, improves over Qwen3-14B on all ten, and gains an average of **2.38 percentage points** over that single-model reference. On the MATH evaluation, CL scores **392/500 correct**, or **78.40% accuracy and 85.25% F1**, under the evaluation scorer.
+
+The exact MultiPL-E 8B checkpoint used for the reported result is unavailable; see [checkpoint availability](#checkpoint-availability) below.
+
+### 📉 Efficiency and task performance
+
+Figure 4 compares task performance with parameter-weighted output cost and average model invocations across the five benchmarks.
+
+<p align="center">
+  <a href="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/Figure4.pdf"><img src="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/previews/Figure4-1.png" alt="Computation cost versus task performance across five benchmarks" width="960"></a>
+</p>
+
+### 🧭 Emergent interaction topology
+
+Figure 5 visualizes the task-dependent interaction topology. The model order changes with the task, allowing CL to draft, review, refine, enrich, or terminate as needed.
+
+<p align="center">
+  <a href="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/Figure5.pdf"><img src="https://huggingface.co/SJTUFreshman/collaborativelearning/resolve/main/assets/figures/previews/Figure5-1.png" alt="Task-dependent learned interaction topology" width="960"></a>
+</p>
+
+## 🧭 Workflow
+
+```text
+task query + current reasoning state
+                  |
+                  v
+          sliced model pool
+                  |
+                  v
+           protocol decision
+          /        |          \
+     continue   handoff     terminate
+        |          |            |
+        +------> next model    v
+                              final answer
+                                   |
+                                   v
+                          task-specific evaluator
+```
+
+At each step, the current task state is processed by one model from the sliced-model pool. The learned protocol can continue the current thread, hand the state to another model, or terminate with a final answer. Protocol distillation provides coordination traces, pyramid supervision scores the reasoning process and final output, and each benchmark supplies a task-specific evaluator.
 
-This repository publishes the surviving rank-64 LoRA checkpoint weights for the heterogeneous collaborative-learning (CL) main experiments in *Collaborative Learning and Reasoning over Model Slices* (paper 338). It covers MuSiQue, GSM-Hard, MATH, MultiPL-E and Conifer, using Qwen3-1.7B, Qwen3-4B and Qwen3-8B. The frozen Qwen base weights are obtained from their original Hugging Face repositories.
+## 📦 Repository layout
 
-The selection is aligned with the supplied paper's Table 1 and Tables 15–17, and the project's final-run training audit, summaries and evaluation provenance. See [PAPER_ALIGNMENT.md](PAPER_ALIGNMENT.md) for the exact run mapping and limitations.
+| Path | Contents |
+| --- | --- |
+| `<benchmark>/PD/<Qwen3-model>/<checkpoint>/` | Protocol distillation (SFT) adapters |
+| `<benchmark>/SLS/<Qwen3-model>/<checkpoint>/` | Step-level supervision (RL) adapters |
+| `assets/figures/` | Study figures and previews |
+| `manifest.json` | Complete checkpoint list, source runs, sizes, SHA-256 hashes, and known missing checkpoints |
+| `SHA256SUMS` | SHA-256 checksums for checkpoint assets |
+| `PAPER_ALIGNMENT.md` | Final training-run mapping and evaluation checkpoint identities |
+| `load_adapter.py` | Helper for loading either adapter layout |
 
-## Scope and layout
+The release contains **233 surviving checkpoints across 29 model-stage groups**, approximately **41.39 GiB** including tokenizer assets and configuration. Each selected final training run retains all its surviving `checkpoint-*` exports and its `final` export. Some checkpoints place the adapter in a nested `policy_adapter/` directory, with the tokenizer in the parent directory. MultiPL-E 1.7B is PD-only.
 
-`<dataset>/<stage>/<base-model>/<checkpoint>/`
+Weights are byte-for-byte unchanged. Adapter configurations use the corresponding `Qwen/Qwen3-*` base-model IDs, and JSON metadata has local server paths normalized and credential fields removed. Baseline, ablation, and homogeneous-model weights are outside this release, as are optimizer/RNG states, pickled training arguments, and training data.
 
-- `PD` is protocol distillation (SFT).
-- `SLS` is step-level supervision / collaborative refinement (named `PS` or `rl_runs` in the source project).
-- `final` contains the export at the end of the selected training run. All other surviving `checkpoint-*` weight exports from the same run are retained.
-- Some checkpoints store the LoRA adapter in a nested `policy_adapter/` directory; the tokenizer remains in its parent checkpoint directory.
-- The exact checkpoint list, source-relative paths, byte sizes and SHA-256 hashes are in [manifest.json](manifest.json). [SHA256SUMS](SHA256SUMS) covers all published checkpoint assets.
-- This is a weight release with adapter configuration, tokenizer assets and recorded JSON training metadata. Optimizer/RNG states, pickled training arguments, training examples, raw logs, baseline models, ablations and homogeneous controls are outside this release.
+### Checkpoint availability
 
-All weight tensor files are byte-for-byte unchanged. Public adapter configurations replace server-local base-model paths with `Qwen/Qwen3-*` model IDs. JSON metadata has server-local paths normalized and credential fields removed.
+The source server's 2026-11-05 cleanup deleted 18 checkpoints from the selected runs. In particular, **MultiPL-E 8B `checkpoint-25`, used in the paper's final evaluation, could not be recovered from either source server**. The same run's `checkpoint-100` and `final` are included as surviving exports; they are not substitutes for checkpoint 25 and are not assigned its reported scores. The manifest lists the known deletions. Earlier retention limits may also have removed checkpoints.
 
-## Availability limitation
+## 🚀 Quickstart
 
-**The exact MultiPL-E 8B `checkpoint-25` used for the paper's 59.39 / 59.63 result is unavailable.** The source server's cleanup record reports its deletion on 2026-11-05. The same run's `checkpoint-100` and `final` survive and are included, but they are not represented as equivalent to checkpoint 25 and are not assigned its evaluation scores.
+### 1️⃣ Install the loading dependencies
 
-The two requested servers were inspected. No additional matching main-experiment weights were found on the Zhongwei server; its other surviving models are homogeneous experiments, baselines or later retraining variants. The manifest records 18 checkpoints from the selected runs that are known to have been removed by the 2026-11-05 cleanup. Earlier training retention may also have removed checkpoints; this repository does not claim to contain every checkpoint ever produced.
+```bash
+python -m pip install -U torch transformers peft accelerate huggingface_hub
+```
 
-## Load a checkpoint
+### 2️⃣ Download and load one adapter
 
 ```python
 from huggingface_hub import snapshot_download
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from peft import PeftModel
 
-repo_id = "SJTUFreshman/JCA-main-experiment-checkpoints"
+repo_id = "SJTUFreshman/collaborativelearning"
 subfolder = "MuSiQue/SLS/Qwen3-1.7B/final"
 snapshot = snapshot_download(repo_id, allow_patterns=[f"{subfolder}/*"])
 adapter_path = f"{snapshot}/{subfolder}"
+
 tokenizer = AutoTokenizer.from_pretrained(adapter_path)
 base_model = AutoModelForCausalLM.from_pretrained(
     "Qwen/Qwen3-1.7B", torch_dtype="auto", device_map="auto"
@@ -58,7 +175,31 @@ base_model = AutoModelForCausalLM.from_pretrained(
 model = PeftModel.from_pretrained(base_model, adapter_path)
 ```
 
-Use the matching base model from the manifest for each adapter. For a nested `policy_adapter` checkpoint, load the adapter from that nested directory and the tokenizer from its parent. [load_adapter.py](load_adapter.py) handles both layouts. This example loads one agent; reproducing the paper requires the multi-agent scheduler, task-specific prompts, inference settings and evaluation protocol from the original project. The paper lists its code at https://anonymous.4open.science/r/collaborativelearning. The recorded Table 1 numbers are paper-reported results, not a new evaluation of this release.
+Choose the matching base model recorded in `manifest.json`. For a nested `policy_adapter` checkpoint, load the adapter from that directory and the tokenizer from its parent; [load_adapter.py](load_adapter.py) handles both layouts. The frozen base-model weights are downloaded from the original Qwen repositories.
+
+To download all published checkpoints:
+
+```bash
+hf download SJTUFreshman/collaborativelearning --local-dir collaborativelearning
+```
+
+### 3️⃣ Reproduce the collaborative evaluation
+
+Loading one adapter does not run the collaborative system. Use the [companion code repository](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c) for the scheduler, task prompts, model endpoints, and benchmark scorers. Its Quickstart includes a CPU-only MATH summary from the bundled evaluation records and a fresh MATH evaluation using three OpenAI-compatible model endpoints. [PAPER_ALIGNMENT.md](PAPER_ALIGNMENT.md) records the final evaluation identities and the MATH answer-selection protocol.
+
+## 🧰 Code and experiment entry points
+
+The following paths are in the companion code repository:
+
+| Area | Entry point |
+| --- | --- |
+| Shared CL utilities and analysis | [reproduction/code/jca/scripts/](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/tree/main/reproduction/code/jca/scripts) |
+| MATH CL training and evaluation | [reproduction/code/jca/experiments/math_specific_sft_rl_v1/](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/tree/main/reproduction/code/jca/experiments/math_specific_sft_rl_v1) |
+| Conifer pipeline | [reproduction/code/jca/conifer_training_hub/](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/tree/main/reproduction/code/jca/conifer_training_hub) |
+| MultiPL-E scoring | [reproduction/code/jca/Code/MultiPL-E/](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/tree/main/reproduction/code/jca/Code/MultiPL-E) |
+| Saved training records | [reproduction/training/](https://github.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/tree/main/reproduction/training) |
+
+A full run may require Python 3.10+, PyTorch/CUDA, Transformers, PEFT, Accelerate, vLLM, language runtimes, and benchmark-specific evaluators. Use the checked-in configurations and logs for seeds, model servers, and output conventions.
 
 ## License
 

assets/figures/Figure1.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:653045b404edd6a592ca14cac3a84c7b61c3956eea141c04b82b24e74657b4bd
+size 376603

assets/figures/Figure2.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15723dd063604b673d68eabd89878e7b92f1fd56ce6618fe969d6f582cffaa06
+size 817550

assets/figures/Figure4.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2df7e18eaa0aedf3fbfc5646c9f947b12549e953ac4637167841486f4af33aff
+size 270674

assets/figures/Figure5.pdf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2ed311851f6a1521239769bd823141b1b07473aa24552fc693ca9c2166c254d
+size 274540

load_adapter.py

@@ -8,7 +8,7 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 
 
 def load_adapter(subfolder, revision=None):
-    repo_id = 'SJTUFreshman/JCA-main-experiment-checkpoints'
+    repo_id = 'SJTUFreshman/collaborativelearning'
     tokenizer_folder = str(Path(subfolder).parent).replace('\\', '/') if subfolder.endswith('/policy_adapter') else subfolder
     snapshot = snapshot_download(repo_id, revision=revision, allow_patterns=[subfolder+'/*', tokenizer_folder+'/tokenizer*', tokenizer_folder+'/chat_template.jinja', tokenizer_folder+'/special_tokens_map.json'])
     adapter_dir = Path(snapshot)/subfolder

manifest.json

@@ -1,7 +1,7 @@
 {
   "schema_version": 1,
   "release_date": "2026-11-10",
-  "repo_id": "SJTUFreshman/JCA-main-experiment-checkpoints",
+  "repo_id": "SJTUFreshman/collaborativelearning",
   "scope": "Paper Table 1 heterogeneous CL main experiments; all surviving checkpoint weights from the final PD/SLS training runs.",
   "license": null,
   "source_servers_checked": [
@@ -12565,5 +12565,67 @@
       "source_relative_path": "sft_runs/multipl_e_8lang_v5_diverse_a3_5ep/A3/checkpoint-40",
       "deleted_at": "2026-11-05T13:15:03.432493+08:00"
     }
+  ],
+  "documentation_files": [
+    {
+      "path": "assets/figures/previews/Figure1-1.png",
+      "size": 537010,
+      "sha256": "8976af442ec2159e8e4db665a27ff4aaed2da4cfadf5f23ba2d5b9bdcf360737",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/previews/Figure1-1.png"
+    },
+    {
+      "path": "assets/figures/previews/Figure2-1.png",
+      "size": 794633,
+      "sha256": "5bc0549352f1730e2ae5c0efd11b6274e4ccad4f5322f5d1a5f56c139bf4a945",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/previews/Figure2-1.png"
+    },
+    {
+      "path": "assets/figures/previews/Figure3-1.png",
+      "size": 192450,
+      "sha256": "80de77311ae407ac0e59444869cf2bd13ae393267d29e76ad32e2bcedd912eb9",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/previews/Figure3-1.png"
+    },
+    {
+      "path": "assets/figures/previews/Figure4-1.png",
+      "size": 390111,
+      "sha256": "6c8b5aecb1ca6b69e9e893b47288eab238c0b156419b7c5ca86277741f6b3a7a",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/previews/Figure4-1.png"
+    },
+    {
+      "path": "assets/figures/previews/Figure5-1.png",
+      "size": 391850,
+      "sha256": "4c3e8fa525115c02b46bca01493dbcabe8f2e55532fa8a6d176f9dbe7998140d",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/previews/Figure5-1.png"
+    },
+    {
+      "path": "assets/figures/Figure1.pdf",
+      "size": 376603,
+      "sha256": "653045b404edd6a592ca14cac3a84c7b61c3956eea141c04b82b24e74657b4bd",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/Figure1.pdf"
+    },
+    {
+      "path": "assets/figures/Figure2.pdf",
+      "size": 817550,
+      "sha256": "15723dd063604b673d68eabd89878e7b92f1fd56ce6618fe969d6f582cffaa06",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/Figure2.pdf"
+    },
+    {
+      "path": "assets/figures/Figure3.pdf",
+      "size": 93932,
+      "sha256": "41a01ef33c073669a985ba9a397ff75109798323277f9da33a285c95762b75fe",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/Figure3.pdf"
+    },
+    {
+      "path": "assets/figures/Figure4.pdf",
+      "size": 270674,
+      "sha256": "2df7e18eaa0aedf3fbfc5646c9f947b12549e953ac4637167841486f4af33aff",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/Figure4.pdf"
+    },
+    {
+      "path": "assets/figures/Figure5.pdf",
+      "size": 274540,
+      "sha256": "d2ed311851f6a1521239769bd823141b1b07473aa24552fc693ca9c2166c254d",
+      "source_url": "https://raw.githubusercontent.com/SJTUFreshman/tmp_x7k9p4r2d8m5c/main/assets/figures/Figure5.pdf"
+    }
   ]
 }