UGB Chapter 17: The Second Round, Head-to-Head Clash
Chapter 17: The Second Round, Head-to-Head Clash
November 3rd, the large lecture hall on the fifth floor of the administration building at Huaxia University.
The preliminary round of the New Talent Cup was more formal than Zuo Cheng had anticipated—seven judges sat on the stage, three from the university and four from external companies, including technical executives and investors. Twelve booths were set up in the auditorium, each with a long table and a whiteboard, and the participating teams' materials and demonstration equipment had been prepared in advance.
Zuo Cheng's team was assigned to the third booth in Group B. Shen Yue arrived early in the morning and set up the booth neatly and efficiently—three display boards, including the team introduction, technical architecture diagram, and business model canvas, were placed behind the table. Fang Ze's embedded demonstration prototype was placed in the center of the table, next to Chen Hao's laptop, with the roadshow animation designed by Shen Yue on standby.
Zhang Lei was wearing a new shirt, and Zuo Cheng noticed that the label on the collar hadn't been cut off yet, so he tore it off for him.
Are you nervous?
"I'm not nervous," Zhang Lei puffed out his chest, then added quietly, "My palms are just a little sweaty."
"That's normal. Once you're on stage, don't think about the judges; just treat it like you're talking to us in the dorm."
Zhang Lei took a deep breath and nodded.
The semi-final rules were that each group had 15 minutes for a presentation followed by 10 minutes for questions from the judges, and the top six teams advanced to the finals based on their scores. Zuocheng's team was drawn seventh, which was right in the middle – they didn't have to be the first to speak, nor did they have to wait until the judges were tired of hearing their presentations.
Ma Hao's "Sharp Core Technology" ranks tenth.
The presentations from the first six groups varied greatly in quality. Some had solid technical skills but presented them in a disjointed manner; others had flashy PowerPoint presentations but were exposed as weak when asked about technical details; one group presented a novel smart agriculture project but lacked a clear commercialization path. Zuo Cheng listened attentively while taking notes on the judges' questioning habits—which judges liked to ask about technical details, which focused on market size, and which were fixated on financial models—to figure out how to present effectively.
The seventh group, it's their turn.
When Zhang Lei stood on the podium, the sweat on his palms had already dried.
Shen Yue's roadshow animation begins with a scene video—a car is driving into a tunnel at 120 kilometers per hour, the car navigation suddenly loses signal, the screen becomes static, and the automatic driving assistance system pops up a warning. The video pauses, and Zhang Lei's voice takes over.
"Every year, more than 12,000 traffic accidents occur on Earth due to communication signal disruptions. This isn't because there are no cell towers, but because existing signal processing algorithms 'crash' in extreme environments. Our job is to prevent that crash."
For the first thirty seconds, the room was completely silent.
Zuo Cheng sat in the first row below the stage and saw at least four judges raise their heads at the same time.
Shen Yue's presentation logic was executed flawlessly—first, the immersive scenario; then, the technical solution; and finally, data verification. Zhang Lei's performance exceeded Zuo Cheng's expectations. He didn't simply recite his prepared remarks, but spoke in a tone somewhere between a speech and a conversation, neither slick nor stiff. When discussing the core algorithm, he translated the technical terms into plain language: "To put it simply, our algorithm isn't a good student who can only solve standard problems, but a master who can adapt to changing circumstances in an exam—the more difficult the problem, the better it performs."
Several judges laughed. One investor drew a circle on his notebook.
Next was Fang Ze's embedded prototype demonstration. Fang Ze didn't go on stage; Zhang Lei operated it from the podium—he pressed a button, and the waveform of the signal processing was displayed in real time on the prototype's screen. As the simulated scene switched from a city to a tunnel and then to a highway, the waveform fluctuated violently and then quickly stabilized, with a visual impact far exceeding that of static charts on a PowerPoint presentation.
The roadshow ended in thirteen minutes.
During the Q&A session, the first question came from an elderly professor with gray hair: "Your algorithm's performance improvement data at low signal-to-noise ratios is very impressive. Could you explain the underlying technical principles?"
This question was originally supposed to be answered by Zuo Cheng, but Zhang Lei, following the pre-arranged division of labor, smoothly transitioned the conversation: "This question will be answered by our technical lead, Zuo Cheng."
Zuo Cheng stood up and succinctly explained the core mechanism of the positive interaction coupling effect—without formulas, purely using logic and analogy. The old professor nodded after listening and asked a follow-up question about a detail regarding convergence, to which Zuo Cheng answered fluently.
The second question came from an investor: "What are your core technological barriers? If Blue Bay Communications or other large companies want to develop their own similar algorithms, how will you compete?"
Zuo Cheng replied, "Two barriers. First, our core algorithm has been patented, and our intellectual property rights are protected by law. Second, the algorithm's performance advantage stems from the coupling effect between the two modules. This effect was discovered unexpectedly and theoretically proven during our research and development process, and the relevant paper is currently being submitted to a top-tier journal on Earth. Simply imitating the algorithm architecture without understanding the coupling mechanism will not produce the same performance."
The investor drew another circle on the notebook.
After the Q&A session, the applause, though not enthusiastic, was genuine. The Zuo Cheng team returned to their booth and sat down. Zhang Lei finally couldn't hold back any longer and whispered, "My legs are weak," before chugging down half a bottle of mineral water.
Fang Ze handed him a pack of tissues to wipe his sweat without any expression.
Zuo Cheng didn't listen much to the next two roadshows; he was waiting for the tenth one.
When Ma Hao took the stage, he was wearing a dark gray suit, and his hair was styled even more meticulously than it had been at the last group meeting. His team consisted of four people: three graduate students from the School of Microelectronics and one PhD student from the School of Computer Science.
"Sharp Chip Technology" also presented its roadshow in a scenario-based manner, focusing on the optimization of communication infrastructure in smart cities. The PPT was very professionally designed, and Ma Hao's demeanor was steady—after all, he is a second-year doctoral student with rich experience in dissertation defense, speaking calmly and clearly with a logical line of reasoning.
But when Zuo Cheng heard the fifth minute, the corner of his mouth twitched slightly.
Ma Hao's core technical solution uses AI models for channel prediction, a completely different approach from Zuo Cheng's adaptive tracking. However, one module in his solution—channel state feature extraction—uses a method highly similar to the framework that Ma Hao "borrowed" from Zuo Cheng's research group.
It wasn't a direct copy; many changes were made. But the basic structure was there, and Zuo Cheng recognized it at a glance.
He didn't make a fuss.
The competition wasn't an internal matter for the research group. The judges weren't familiar with that period of history, and pointing it out on the spot would only make them seem petty. Besides, Ma Hao had made enough changes; strictly speaking, it didn't count as plagiarism, but rather as "inspired" by the work.
The judges will decide the merits of the technical solutions.
During the Q&A session, someone asked Ma Hao a question about the amount of training data required for AI models. Ma Hao answered satisfactorily, but Zuo Cheng noticed that he avoided a crucial point—the model's generalization ability in small sample scenarios. This is precisely the biggest weakness of AI methods compared to adaptive methods.
The judges didn't press the matter. But Zuo Cheng knew that this question would definitely be brought up in the finals.
After all twelve teams finished their presentations, the judges left for a discussion. The participating teams then freely exchanged ideas in the exhibition area, and the atmosphere became much more relaxed.
Ma Hao walked out of the crowd and stopped in front of Zuo Cheng.
"The roadshow went well," he said with a smile, his tone polite and courteous. "Your embedded demo prototype was quite interesting. Where was it made?"
"The team members did it themselves." Zuo Cheng smiled, equally polite.
The two stared at each other for a second, and there was something in the air that only they could sense. It wasn't hostility, but a cold, knowing understanding—I know what you're doing, and you know I know.
"See you in the finals," Ma Hao said.
"See you in the finals," Zuo Cheng replied.
Twenty minutes later, the results of the semi-finals were announced.
The six teams that advanced to the finals from Group B were announced one by one. Zuo Cheng's team ranked second, and Ma Hao's "Sharp Core Technology" ranked fifth.
Zhang Lei almost jumped up behind the booth, but Chen Hao grabbed his shoulder and held him down.
Shen Yue took out her phone, snapped a picture of the results announcement, and sent it to the team group chat. Liu Wei immediately replied with a single word: "Steady!"
Zuo Cheng stood in front of the booth, looking at the final date posted on the wall—November 17th.
Two more weeks.
Two weeks later, the Rising Star Cup finals and the final stage of the second phase of the main quest chain will arrive almost simultaneously.
Two battles must be fought simultaneously.
He stretched his neck and picked up the bottled water on the table, taking a sip.
Not afraid.
He was never afraid to fight two battles at the same time.