Further Reading

A curated, deliberately short list of resources that have held up. Links are plain text so this library stays self-contained and offline-friendly — copy a URL into a browser when you're online. This is not exhaustive; it's a set of durable starting points, and the fast-moving areas are flagged as such.

Foundational papers

These are the primary sources behind much of the Foundations section. Dense but rewarding.

• Attention Is All You Need (Vaswani et al., 2017) — introduces the transformer. The origin point for modern LLMs. https://arxiv.org/abs/1706.03762
• BERT: Pre-training of Deep Bidirectional Transformers (Devlin et al., 2018) — the encoder-only, bidirectional approach; foundational for embeddings and classification. https://arxiv.org/abs/1810.04805
• Language Models are Few-Shot Learners (Brown et al., 2020) — the GPT-3 paper; established in-context / few-shot learning at scale. https://arxiv.org/abs/2005.14165
• Training language models to follow instructions with human feedback (Ouyang et al., 2022) — InstructGPT; the basis of modern instruction tuning and RLHF. https://arxiv.org/abs/2203.02155

Retrieval and RAG

• Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (Lewis et al., 2020) — the paper that named RAG. https://arxiv.org/abs/2005.11401
• Dense Passage Retrieval for Open-Domain QA (Karpukhin et al., 2020) — dense embedding-based retrieval. https://arxiv.org/abs/2004.04906

Prompting and reasoning

• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models (Wei et al., 2022) — the chain-of-thought paper. https://arxiv.org/abs/2201.11903
• Self-Consistency Improves Chain of Thought Reasoning (Wang et al., 2022) — sample-and-vote over multiple reasoning chains. https://arxiv.org/abs/2203.11171

Agents

• ReAct: Synergizing Reasoning and Acting in Language Models (Yao et al., 2022) — the reason-act loop. https://arxiv.org/abs/2210.03629
• Toolformer: Language Models Can Teach Themselves to Use Tools (Schick et al., 2023) — an early take on models learning tool use. https://arxiv.org/abs/2302.04761

Safety and evaluation

• OWASP Top 10 for LLM Applications — a practitioner's catalog of LLM security risks, including prompt injection. Updated periodically; find the current version. https://owasp.org/www-project-top-10-for-large-language-model-applications/
• NIST AI Risk Management Framework — a structured way to think about AI risk. https://www.nist.gov/itl/ai-risk-management-framework
• Judging LLM-as-a-Judge (MT-Bench / Chatbot Arena) (Zheng et al., 2023) — foundational study of LLM-as-judge and its biases. https://arxiv.org/abs/2306.05685

Living resources (fast-moving — expect churn)

These change frequently; treat them as pointers, not fixed truth.

• Provider documentation — OpenAI, Anthropic, Google, Meta, Mistral, and Ollama each publish current guides on prompting, tools, and structured output. For anything about a specific model's limits, pricing, or features, the provider's own docs are the only authority — this library deliberately avoids quoting numbers that go stale.
• Model cards — read the card for any embedding or chat model before relying on it; usage conventions (query prefixes, symmetric vs. asymmetric) live there.
• Community leaderboards and eval suites — useful for orientation, but no substitute for your own evaluation on your own task.

How to use this list

Read the foundational papers once for real understanding; keep the living resources bookmarked for current specifics. And remember the recurring theme of this whole library: where a topic is fast-moving, verify against a current primary source rather than trusting any summary — including this one.

Return to the Reference section or the library home.