Essential Papers¶
Curated papers every AI engineer should know — not an exhaustive dump. For the full auto-index see Papers.
Foundations¶
| Paper | Why it matters | Module |
|---|---|---|
| Attention Is All You Need (Vaswani et al.) | Transformer architecture | M00, M06 |
| BERT | Bidirectional pre-training | M00 |
| Language Models are Few-Shot Learners (GPT-3) | In-context learning | M07 |
| Training language models to follow instructions (InstructGPT) | Instruction tuning | M07 |
RAG & retrieval¶
| Paper | Why it matters | Module |
|---|---|---|
| Retrieval-Augmented Generation | RAG foundation | M09 |
| Dense Passage Retrieval | Dense retrieval | M09, M13 |
Agents¶
| Paper | Why it matters | Module |
|---|---|---|
| ReAct | Reason + act loop | M11 |
| Toolformer | Self-supervised tool use | M11, M18 |
| Generative Agents | Multi-agent simulation | M12 |
Evaluation & safety¶
| Paper | Why it matters | Module |
|---|---|---|
| HELM | Holistic LLM benchmarking | M19 |
| Judging LLM-as-a-Judge (MT-Bench) | LLM judges | M19 |
| Constitutional AI | RLHF alternative | M16 |
Visual companion¶
- The Illustrated Transformer — Jay Alammar (open educational)