Department of Computer Engineering · Akdeniz Üniversitesi
LLM Research
Group
A research group at Akdeniz Üniversitesi working on large language models.
Attention(Q, K, V) = softmax(QKᵀ / √dk) V
decoder-only · greedy decoding00/12 tokens
Prompt
Summarise in one sentence: self-attention lets every position in a sequence look at every earlier position.
Completion
Next-token distribution
- Each0.61
- Every0.22
- Tokens0.08
Scripted illustration of next-token prediction — not a live model.
- Publications
- 03
- Group members
- 07
- Doctoral students
- 05
- Alumni
- 00
§ 01 — Research
Work currently underway
The group is new. Our research is best described by the doctoral theses in progress.
The decoder block our work sits inside — activations flow up the residual stream, N layers deep.
- 01Natural language processing, data science and large language modelsMelih Günay · Prof. Dr.
- 02Machine and deep learning, computer vision and image processingHüseyin Gökhan Akçay · Asst. Prof. Dr.
- 03Clinical Text SummarizationMuhammad Sanwal · PhD Student
- 04Multi-agent reasoningSeth Ampofo · PhD Student
- 05Transport optimization using large language modelsBarış Doruk Başaran · PhD Student
