Jerry commited on
Commit
c57da4f
ยท
1 Parent(s): 95b0b90

feat: add 9-slide HTML presentation deck and link in README

Browse files
Files changed (2) hide show
  1. README.md +3 -1
  2. slides.html +375 -0
README.md CHANGED
@@ -226,9 +226,10 @@ logistics-hackathon-env/
226
  | Resource | Link |
227
  |----------|------|
228
  | ๐Ÿค— Live HF Space | [huggingface.co/spaces/Leavin1611/logistics-hackathon-env](https://huggingface.co/spaces/Leavin1611/logistics-hackathon-env) |
 
229
  | ๐Ÿ““ Training Notebook | [train_colab.ipynb (Colab)](https://colab.research.google.com/drive/1fRfheRZd1tffKjXKZkW72ffl9JPxDpL8) |
 
230
  | ๐Ÿ“ Mini-Blog | [HF_BLOG_POST.md](./HF_BLOG_POST.md) |
231
- | ๐Ÿง  Trained Model | [Leavin1611/logistics-hackathon-model](https://huggingface.co/Leavin1611/logistics-hackathon-model) |
232
  | ๐Ÿ“ฆ GitHub Repo | [github.com/leavin1611/Logistics-hackathon-env](https://github.com/leavin1611/Logistics-hackathon-env) |
233
  | ๐Ÿ”ง OpenEnv Framework | [github.com/meta-pytorch/OpenEnv](https://github.com/meta-pytorch/OpenEnv) |
234
  | ๐Ÿ”‘ Free Groq API Key | [console.groq.com/keys](https://console.groq.com/keys) |
@@ -238,6 +239,7 @@ logistics-hackathon-env/
238
  ## ๐Ÿ“š Further Reading
239
 
240
  - [HF_BLOG_POST.md](./HF_BLOG_POST.md) โ€” Full narrative writeup for judges
 
241
  - [DESIGN.md](./DESIGN.md) โ€” Architecture decisions and reward anatomy
242
  - [CONTRIBUTING.md](./CONTRIBUTING.md) โ€” How to add scenarios, routes, actions
243
 
 
226
  | Resource | Link |
227
  |----------|------|
228
  | ๐Ÿค— Live HF Space | [huggingface.co/spaces/Leavin1611/logistics-hackathon-env](https://huggingface.co/spaces/Leavin1611/logistics-hackathon-env) |
229
+ | ๐Ÿง  Trained Model | [Leavin1611/logistics-hackathon-model](https://huggingface.co/Leavin1611/logistics-hackathon-model) |
230
  | ๐Ÿ““ Training Notebook | [train_colab.ipynb (Colab)](https://colab.research.google.com/drive/1fRfheRZd1tffKjXKZkW72ffl9JPxDpL8) |
231
+ | ๐Ÿ–ฅ๏ธ Slide Deck | [slides.html](./slides.html) โ€” 9-slide presentation (use arrow keys to navigate) |
232
  | ๐Ÿ“ Mini-Blog | [HF_BLOG_POST.md](./HF_BLOG_POST.md) |
 
233
  | ๐Ÿ“ฆ GitHub Repo | [github.com/leavin1611/Logistics-hackathon-env](https://github.com/leavin1611/Logistics-hackathon-env) |
234
  | ๐Ÿ”ง OpenEnv Framework | [github.com/meta-pytorch/OpenEnv](https://github.com/meta-pytorch/OpenEnv) |
235
  | ๐Ÿ”‘ Free Groq API Key | [console.groq.com/keys](https://console.groq.com/keys) |
 
239
  ## ๐Ÿ“š Further Reading
240
 
241
  - [HF_BLOG_POST.md](./HF_BLOG_POST.md) โ€” Full narrative writeup for judges
242
+ - [slides.html](./slides.html) โ€” 9-slide presentation deck (arrow key navigation)
243
  - [DESIGN.md](./DESIGN.md) โ€” Architecture decisions and reward anatomy
244
  - [CONTRIBUTING.md](./CONTRIBUTING.md) โ€” How to add scenarios, routes, actions
245
 
slides.html ADDED
@@ -0,0 +1,375 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!DOCTYPE html>
2
+ <html lang="en">
3
+ <head>
4
+ <meta charset="UTF-8">
5
+ <meta name="viewport" content="width=device-width, initial-scale=1.0">
6
+ <title>AI Logistics Coordinator โ€” OpenEnv Hackathon 2026</title>
7
+ <style>
8
+ @import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;600;700;900&family=JetBrains+Mono:wght@400;600&display=swap');
9
+
10
+ :root {
11
+ --bg: #0d1117; --surface: #161b22; --border: #30363d;
12
+ --text: #e6edf3; --muted: #8b949e; --green: #3fb950;
13
+ --red: #f85149; --blue: #58a6ff; --orange: #f0883e;
14
+ --accent: #238636;
15
+ }
16
+
17
+ * { margin: 0; padding: 0; box-sizing: border-box; }
18
+ html, body { height: 100%; background: var(--bg); font-family: 'Inter', sans-serif; color: var(--text); overflow: hidden; }
19
+
20
+ .deck { width: 100%; height: 100vh; position: relative; }
21
+
22
+ .slide {
23
+ display: none; position: absolute; inset: 0;
24
+ flex-direction: column; justify-content: center; align-items: center;
25
+ padding: 60px 80px; text-align: center;
26
+ background: radial-gradient(ellipse at 50% 0%, #161b22 0%, #0d1117 70%);
27
+ }
28
+ .slide.active { display: flex; }
29
+
30
+ h1 { font-size: 2.8rem; font-weight: 900; line-height: 1.15; }
31
+ h2 { font-size: 2rem; font-weight: 700; margin-bottom: 1rem; }
32
+ h3 { font-size: 1.3rem; font-weight: 600; color: var(--blue); margin-bottom: 0.5rem; }
33
+ p { font-size: 1.1rem; line-height: 1.7; color: var(--muted); max-width: 800px; }
34
+
35
+ .tag {
36
+ display: inline-block; padding: 4px 12px; border-radius: 20px;
37
+ font-size: 0.8rem; font-weight: 600; margin: 4px;
38
+ background: rgba(88,166,255,0.1); border: 1px solid var(--blue); color: var(--blue);
39
+ }
40
+ .tag.green { background: rgba(63,185,80,0.1); border-color: var(--green); color: var(--green); }
41
+ .tag.orange { background: rgba(240,136,62,0.1); border-color: var(--orange); color: var(--orange); }
42
+
43
+ .highlight {
44
+ font-size: 4rem; font-weight: 900;
45
+ background: linear-gradient(135deg, var(--green), var(--blue));
46
+ -webkit-background-clip: text; -webkit-text-fill-color: transparent;
47
+ }
48
+ .sub { font-size: 1rem; color: var(--muted); margin-top: 0.25rem; }
49
+
50
+ .grid { display: grid; grid-template-columns: repeat(3, 1fr); gap: 20px; width: 100%; max-width: 1000px; margin-top: 1.5rem; }
51
+ .grid-2 { grid-template-columns: repeat(2, 1fr); }
52
+ .grid-4 { grid-template-columns: repeat(4, 1fr); }
53
+
54
+ .card {
55
+ background: var(--surface); border: 1px solid var(--border); border-radius: 12px;
56
+ padding: 20px; text-align: left;
57
+ }
58
+ .card .icon { font-size: 1.8rem; margin-bottom: 8px; }
59
+ .card h4 { font-size: 1rem; font-weight: 600; margin-bottom: 6px; }
60
+ .card p { font-size: 0.85rem; color: var(--muted); max-width: none; }
61
+
62
+ .metric { display: flex; flex-direction: column; align-items: center; padding: 24px 12px; }
63
+ .metric .num { font-size: 2.4rem; font-weight: 900; color: var(--green); }
64
+ .metric .label { font-size: 0.8rem; color: var(--muted); text-align: center; margin-top: 4px; }
65
+
66
+ .badge {
67
+ display: inline-block; padding: 6px 16px; border-radius: 6px;
68
+ background: var(--accent); color: #fff; font-weight: 700;
69
+ font-size: 0.9rem; margin: 6px;
70
+ }
71
+ .badge.red { background: #6e1d1a; }
72
+ .badge.blue { background: #0c2d6b; }
73
+
74
+ table { width: 100%; border-collapse: collapse; font-size: 0.9rem; margin-top: 1rem; }
75
+ th { background: var(--surface); color: var(--blue); padding: 10px 14px; text-align: left; border-bottom: 1px solid var(--border); }
76
+ td { padding: 10px 14px; border-bottom: 1px solid var(--border); color: var(--muted); }
77
+ td strong { color: var(--text); }
78
+
79
+ .reward-bar { display: flex; align-items: center; gap: 12px; margin: 6px 0; }
80
+ .reward-bar .label { width: 160px; text-align: right; font-size: 0.85rem; color: var(--muted); flex-shrink: 0; }
81
+ .reward-bar .bar-wrap { flex: 1; background: var(--border); border-radius: 4px; height: 16px; }
82
+ .reward-bar .bar { height: 16px; border-radius: 4px; transition: width 0.6s ease; }
83
+ .reward-bar .val { font-size: 0.85rem; font-weight: 700; flex-shrink: 0; width: 55px; }
84
+
85
+ .code {
86
+ font-family: 'JetBrains Mono', monospace; font-size: 0.85rem;
87
+ background: var(--surface); border: 1px solid var(--border);
88
+ border-radius: 8px; padding: 16px; text-align: left;
89
+ color: var(--green); width: 100%; max-width: 700px; margin-top: 1rem;
90
+ }
91
+ .code .comment { color: var(--muted); }
92
+
93
+ .penalty { color: var(--red); font-weight: 700; }
94
+ .bonus { color: var(--green); font-weight: 700; }
95
+
96
+ /* Nav */
97
+ .nav {
98
+ position: fixed; bottom: 30px; left: 50%; transform: translateX(-50%);
99
+ display: flex; gap: 12px; align-items: center; z-index: 100;
100
+ background: rgba(22,27,34,0.9); border: 1px solid var(--border);
101
+ border-radius: 40px; padding: 10px 20px; backdrop-filter: blur(10px);
102
+ }
103
+ .nav button {
104
+ background: none; border: none; color: var(--muted); cursor: pointer;
105
+ font-size: 1.1rem; padding: 4px 10px; border-radius: 20px; transition: all 0.2s;
106
+ }
107
+ .nav button:hover { background: var(--surface); color: var(--text); }
108
+ .nav .counter { color: var(--muted); font-size: 0.85rem; min-width: 50px; text-align: center; }
109
+
110
+ .slide-num {
111
+ position: fixed; top: 20px; right: 30px;
112
+ font-size: 0.8rem; color: var(--border);
113
+ }
114
+
115
+ .logo { font-size: 3.5rem; margin-bottom: 16px; }
116
+ .subtitle { font-size: 1.2rem; color: var(--muted); margin-top: 12px; }
117
+ .spacer { margin-top: 1.5rem; }
118
+
119
+ .link { color: var(--blue); text-decoration: none; font-size: 0.85rem; }
120
+ .link:hover { text-decoration: underline; }
121
+
122
+ /* Animate in */
123
+ .slide.active .card, .slide.active .metric {
124
+ animation: fadeUp 0.4s ease both;
125
+ }
126
+ .slide.active .card:nth-child(2) { animation-delay: 0.08s; }
127
+ .slide.active .card:nth-child(3) { animation-delay: 0.16s; }
128
+ .slide.active .card:nth-child(4) { animation-delay: 0.24s; }
129
+
130
+ @keyframes fadeUp {
131
+ from { opacity: 0; transform: translateY(20px); }
132
+ to { opacity: 1; transform: translateY(0); }
133
+ }
134
+
135
+ .arrow { font-size: 1.4rem; color: var(--border); margin: 0 8px; }
136
+ .flow { display: flex; align-items: center; justify-content: center; flex-wrap: wrap; margin-top: 1.5rem; gap: 4px; }
137
+ .flow-step {
138
+ background: var(--surface); border: 1px solid var(--border); border-radius: 8px;
139
+ padding: 10px 18px; font-size: 0.9rem; font-weight: 600;
140
+ }
141
+ </style>
142
+ </head>
143
+ <body>
144
+ <div class="deck">
145
+
146
+ <!-- SLIDE 1: TITLE -->
147
+ <div class="slide active">
148
+ <div class="logo">๐Ÿš›</div>
149
+ <h1>AI Logistics Coordinator</h1>
150
+ <p class="subtitle">Teaching an LLM to manage freight crises using GRPO reinforcement learning</p>
151
+ <div class="spacer"></div>
152
+ <div>
153
+ <span class="tag green">OpenEnv Compatible</span>
154
+ <span class="tag">GRPO + TRL + Unsloth</span>
155
+ <span class="tag orange">Qwen2.5-1.5B</span>
156
+ <span class="tag">Meta PyTorch Hackathon 2026</span>
157
+ </div>
158
+ <div class="spacer"></div>
159
+ <p>
160
+ <a class="link" href="https://huggingface.co/spaces/Leavin1611/logistics-hackathon-env">๐Ÿค— Live Space</a> &nbsp;ยท&nbsp;
161
+ <a class="link" href="https://huggingface.co/Leavin1611/logistics-hackathon-model">๐Ÿง  Trained Model</a> &nbsp;ยท&nbsp;
162
+ <a class="link" href="https://colab.research.google.com/drive/1fRfheRZd1tffKjXKZkW72ffl9JPxDpL8">๐Ÿ““ Colab Notebook</a>
163
+ </p>
164
+ </div>
165
+
166
+ <!-- SLIDE 2: THE PROBLEM -->
167
+ <div class="slide">
168
+ <h2>It's 3 AM. A port strike just shut down Mumbai.</h2>
169
+ <p>COVID vaccines are stranded on Route R1. Election ballots are stuck at Delhi. โ‚น1.8 crore of server hardware is 7 hours late. A human dispatcher needs 30 minutes to triage this. We built an agent that does it in seconds.</p>
170
+ <div class="grid grid-2" style="margin-top: 2rem; max-width: 800px;">
171
+ <div class="card">
172
+ <div class="icon">โŒ</div>
173
+ <h4>Current Systems</h4>
174
+ <p>Rule-based optimizers. Great at routine scheduling. Fail completely when disruptions cascade. Cannot reason, cannot prioritize, cannot communicate.</p>
175
+ </div>
176
+ <div class="card">
177
+ <div class="icon">โœ…</div>
178
+ <h4>Our Agent</h4>
179
+ <p>Reasons about network congestion, triages by urgency, reroutes strategically, and sends empathetic messages to customers โ€” all in a single turn.</p>
180
+ </div>
181
+ </div>
182
+ </div>
183
+
184
+ <!-- SLIDE 3: ENVIRONMENT -->
185
+ <div class="slide">
186
+ <h2>๐ŸŒ The Environment</h2>
187
+ <p>A real-time Indian freight network crisis simulator. The agent plays a centralized logistics coordinator managing cascading disruptions across JNPT, Delhi, and Mundra.</p>
188
+ <div class="grid" style="margin-top: 1.5rem;">
189
+ <div class="card">
190
+ <div class="icon">๐Ÿ‘๏ธ</div>
191
+ <h4>What Agent Sees</h4>
192
+ <p>Shipment states, route congestion (0โ€“100%), active disruptions, SLA time remaining, carrier availability</p>
193
+ </div>
194
+ <div class="card">
195
+ <div class="icon">๐ŸŽฎ</div>
196
+ <h4>What Agent Does</h4>
197
+ <p><code>get_network_status</code> ยท <code>reroute_shipment</code> ยท <code>set_priority</code> ยท <code>communicate_eta</code> ยท <code>end_turn</code></p>
198
+ </div>
199
+ <div class="card">
200
+ <div class="icon">๐Ÿ†</div>
201
+ <h4>Three Task Tiers</h4>
202
+ <p>EASY (2 ships, 1 disruption) โ†’ MEDIUM (4 ships, 3 disruptions) โ†’ HARD (7 ships, 4 failures)</p>
203
+ </div>
204
+ </div>
205
+ </div>
206
+
207
+ <!-- SLIDE 4: WHAT MAKES IT NOVEL -->
208
+ <div class="slide">
209
+ <h2>๐Ÿ”ฌ What Makes This Novel</h2>
210
+ <div class="grid grid-2" style="max-width: 900px; margin-top: 1.5rem;">
211
+ <div class="card">
212
+ <div class="icon">๐ŸŒ</div>
213
+ <h4>Multi-Agent Pressure</h4>
214
+ <p>Background agents compete for the same limited route capacity every turn. The agent must model the network โ€” not just react to individual shipments.</p>
215
+ </div>
216
+ <div class="card">
217
+ <div class="icon">โณ</div>
218
+ <h4>Long-Horizon Planning</h4>
219
+ <p>5โ€“7 turn episodes. Early decisions cascade. Saving the pharmaceutical shipment on turn 1 changes which routes are available on turn 4.</p>
220
+ </div>
221
+ <div class="card">
222
+ <div class="icon">๐Ÿง </div>
223
+ <h4>World Modeling Required</h4>
224
+ <p>Partially observable. The agent must call tools to see the network. It cannot assume anything โ€” it must build its own world model each turn.</p>
225
+ </div>
226
+ <div class="card">
227
+ <div class="icon">๐Ÿ’ฌ</div>
228
+ <h4>NLP-Graded Communication</h4>
229
+ <p>ETA messages are scored for empathy, specificity, and cause explanation. A unique test of "soft skills" that no other logistics env tests.</p>
230
+ </div>
231
+ </div>
232
+ </div>
233
+
234
+ <!-- SLIDE 5: REWARD DESIGN -->
235
+ <div class="slide">
236
+ <h2>๐ŸŽฏ Anti-Gameable Reward Design</h2>
237
+ <p>Three independent reward functions. Each has explicit negative penalties to prevent exploitation.</p>
238
+ <div class="grid" style="margin-top: 1.5rem;">
239
+ <div class="card">
240
+ <h4>๐Ÿ—๏ธ Structure Reward</h4>
241
+ <p><span class="bonus">+0.4</span> Start with status check<br><span class="bonus">+0.3</span> End with end_turn<br><span class="penalty">-0.5</span> Duplicate end_turn (spam)<br><span class="penalty">-0.3</span> Repeated status calls</p>
242
+ </div>
243
+ <div class="card">
244
+ <h4>๐Ÿ›ฃ๏ธ Routing Reward</h4>
245
+ <p><span class="bonus">+0.9</span> Clear route chosen<br><span class="bonus">+0.5</span> All ships rerouted<br><span class="penalty">-0.6</span> Congested route<br><span class="penalty">-0.8</span> Empty route ID (loophole)</p>
246
+ </div>
247
+ <div class="card">
248
+ <h4>๐Ÿ“ข Communication Reward</h4>
249
+ <p><span class="bonus">+0.4</span> Message to delayed ship<br><span class="bonus">+0.2</span> Empathy keywords<br><span class="bonus">+0.1</span> Specific ETA given<br><span class="penalty">-0.3</span> Duplicate message (spam)</p>
250
+ </div>
251
+ </div>
252
+ </div>
253
+
254
+ <!-- SLIDE 6: TRAINING PIPELINE -->
255
+ <div class="slide">
256
+ <h2>๐Ÿ‹๏ธ 4-Phase GRPO Curriculum</h2>
257
+ <p>Group Relative Policy Optimization โ€” the same algorithm behind DeepSeek-R1 โ€” on a free Colab T4 GPU.</p>
258
+ <div class="flow">
259
+ <div class="flow-step">๐ŸŸข Phase 1<br><small>TASK-EASY<br>lr=2e-5</small></div>
260
+ <span class="arrow">โ†’</span>
261
+ <div class="flow-step">๐ŸŸก Phase 2<br><small>TASK-MEDIUM<br>lr=1e-5</small></div>
262
+ <span class="arrow">โ†’</span>
263
+ <div class="flow-step">๐Ÿ”ฅ Phase 3<br><small>Mixed Hardening<br>6 rollouts</small></div>
264
+ <span class="arrow">โ†’</span>
265
+ <div class="flow-step">๐Ÿšจ Phase 4<br><small>Correction Cycle<br>Loophole patch</small></div>
266
+ </div>
267
+ <div class="grid grid-4" style="margin-top: 1.5rem; max-width: 900px;">
268
+ <div class="metric"><div class="num">1.5B</div><div class="label">Model size<br>(Qwen2.5)</div></div>
269
+ <div class="metric"><div class="num">4-bit</div><div class="label">QLoRA quantization<br>(Unsloth)</div></div>
270
+ <div class="metric"><div class="num">150+</div><div class="label">Training<br>episodes</div></div>
271
+ <div class="metric"><div class="num">Free</div><div class="label">Colab T4<br>GPU</div></div>
272
+ </div>
273
+ </div>
274
+
275
+ <!-- SLIDE 7: RESULTS -->
276
+ <div class="slide">
277
+ <h2>๐Ÿ“ˆ Results: Observable Evidence</h2>
278
+ <div class="highlight">+327%</div>
279
+ <div class="sub">Live environment reward improvement (0.18 baseline โ†’ 0.7683 trained)</div>
280
+ <div style="width: 100%; max-width: 700px; margin-top: 2rem;">
281
+ <div class="reward-bar">
282
+ <div class="label">Untrained Baseline</div>
283
+ <div class="bar-wrap"><div class="bar" style="width:23.4%;background:var(--red)"></div></div>
284
+ <div class="val" style="color:var(--red)">0.18</div>
285
+ </div>
286
+ <div class="reward-bar">
287
+ <div class="label">After Phase 1 (Easy)</div>
288
+ <div class="bar-wrap"><div class="bar" style="width:54.7%;background:var(--muted)"></div></div>
289
+ <div class="val" style="color:var(--muted)">0.42</div>
290
+ </div>
291
+ <div class="reward-bar">
292
+ <div class="label">After Phase 2 (Medium)</div>
293
+ <div class="bar-wrap"><div class="bar" style="width:79.4%;background:var(--orange)"></div></div>
294
+ <div class="val" style="color:var(--orange)">0.61</div>
295
+ </div>
296
+ <div class="reward-bar">
297
+ <div class="label">Final (lora-final-perfect)</div>
298
+ <div class="bar-wrap"><div class="bar" style="width:100%;background:var(--green)"></div></div>
299
+ <div class="val" style="color:var(--green)">0.7683</div>
300
+ </div>
301
+ </div>
302
+ </div>
303
+
304
+ <!-- SLIDE 8: WHAT THE AGENT LEARNED -->
305
+ <div class="slide">
306
+ <h2>๐Ÿค– What the Agent Actually Learned</h2>
307
+ <div class="grid grid-2" style="max-width: 900px; margin-top: 1.5rem;">
308
+ <div class="card">
309
+ <div class="icon">๐Ÿ”</div>
310
+ <h4>World Modeling</h4>
311
+ <p>Always calls <code>get_network_status</code> first. Builds a mental model of the entire network before acting โ€” not guessing.</p>
312
+ </div>
313
+ <div class="card">
314
+ <div class="icon">๐Ÿ›ฃ๏ธ</div>
315
+ <h4>Strategic Routing</h4>
316
+ <p>Checks route congestion before rerouting. Avoids the -0.6 penalty for choosing overloaded routes. 3+ reroutes per episode vs 1 before training.</p>
317
+ </div>
318
+ <div class="card">
319
+ <div class="icon">๐Ÿ’ฌ</div>
320
+ <h4>Empathetic Communication</h4>
321
+ <p>Sends well-formed apologies with specific ETAs and causes. "We sincerely apologise โ€” your shipment will arrive by 6PM due to port congestion."</p>
322
+ </div>
323
+ <div class="card">
324
+ <div class="icon">๐Ÿšซ</div>
325
+ <h4>Stopped Hacking</h4>
326
+ <p>Stopped spamming messages and empty routes after Phase 4 correction. Learned that strategic play earns more reward than loopholes.</p>
327
+ </div>
328
+ </div>
329
+ </div>
330
+
331
+ <!-- SLIDE 9: LINKS & RESOURCES -->
332
+ <div class="slide">
333
+ <h2>๐Ÿ”— All Submission Links</h2>
334
+ <table style="max-width: 700px; margin-top: 1rem;">
335
+ <tr><th>Resource</th><th>Link</th></tr>
336
+ <tr><td>๐Ÿค— Live HF Space</td><td><a class="link" href="https://huggingface.co/spaces/Leavin1611/logistics-hackathon-env">spaces/Leavin1611/logistics-hackathon-env</a></td></tr>
337
+ <tr><td>๐Ÿง  Trained Model</td><td><a class="link" href="https://huggingface.co/Leavin1611/logistics-hackathon-model">Leavin1611/logistics-hackathon-model</a></td></tr>
338
+ <tr><td>๐Ÿ““ Training Notebook</td><td><a class="link" href="https://colab.research.google.com/drive/1fRfheRZd1tffKjXKZkW72ffl9JPxDpL8">Open in Colab</a></td></tr>
339
+ <tr><td>๐Ÿ“ Mini-Blog</td><td><a class="link" href="https://huggingface.co/spaces/Leavin1611/logistics-hackathon-env/blob/main/HF_BLOG_POST.md">HF_BLOG_POST.md</a></td></tr>
340
+ <tr><td>๐Ÿ—๏ธ Environment Code</td><td><a class="link" href="https://huggingface.co/spaces/Leavin1611/logistics-hackathon-env/blob/main/server/environment.py">server/environment.py</a></td></tr>
341
+ </table>
342
+ <div class="spacer"></div>
343
+ <p><em>Stack: OpenEnv ยท FastAPI ยท TRL ยท GRPO ยท Unsloth ยท Qwen2.5-1.5B ยท Google Colab T4</em></p>
344
+ </div>
345
+
346
+ </div>
347
+
348
+ <!-- Nav -->
349
+ <div class="nav">
350
+ <button onclick="prev()">โ—€</button>
351
+ <span class="counter" id="counter">1 / 9</span>
352
+ <button onclick="next()">โ–ถ</button>
353
+ </div>
354
+
355
+ <script>
356
+ let current = 0;
357
+ const slides = document.querySelectorAll('.slide');
358
+
359
+ function show(n) {
360
+ slides[current].classList.remove('active');
361
+ current = (n + slides.length) % slides.length;
362
+ slides[current].classList.add('active');
363
+ document.getElementById('counter').textContent = `${current+1} / ${slides.length}`;
364
+ }
365
+
366
+ function next() { show(current + 1); }
367
+ function prev() { show(current - 1); }
368
+
369
+ document.addEventListener('keydown', e => {
370
+ if (e.key === 'ArrowRight' || e.key === ' ') next();
371
+ if (e.key === 'ArrowLeft') prev();
372
+ });
373
+ </script>
374
+ </body>
375
+ </html>