Jun 9, 2026 · resources · 24 min read · 4800 words

52 weeks of AI papers — the 2024 reading list.

resources ml llm 2024 reading-list

One landmark 2024 paper per week — 52 picks, curated down from the year's flood, ordered as a year-long reading plan. Not every model release, not every benchmark: the papers that mattered, the ones the 2025 reasoning explosion was built on (GRPO, test-time scaling, MoE, latent reasoning) plus the architecture bets that paid off later (Mamba-2, BitNet, KAN, byte-level). Read one a week and you walk into 2025 knowing exactly how the field got there.

2024 was the year the modern LLM stack solidified. Open mixture-of-experts went mainstream, RL post-training got a real algorithm (GRPO), test-time compute became a scaling axis, and a dozen teams quietly shipped the architecture ideas that 2025 and 2026 would run with.

There were thousands of papers. A flat list helps nobody. So this is 52 — one per week — picked for impact and for how well they explain the year. I leaned on the famous public 2024 roundups for coverage, then cut hard and added the papers I think actually moved the needle. If you read this list a week at a time, by week 52 you understand 2024 well enough that the 2025–2026 list reads like a sequel you already know the plot of.

How dates work: the date is the first arXiv submission month. The id (e.g. 2402.03300) encodes year-month — 24=2024, 02=February. Every row keeps the id so you can verify and cite. Links go to the /abs/ page, so you land on versions, code, and BibTeX.

How to use this list

As a calendar. One paper a week, in order. The weeks are roughly chronological, so you experience 2024 as it unfolded — the MoE wave, the architecture experiments, the model-report summer, the reasoning-and-efficiency finale in December.

As a reference. Skim the "why read" column, jump to whatever theme you need. The emoji tags let you triage in seconds.

As prep for 2025. Anything tagged 🏛️ is load-bearing — those are the papers the next year cites constantly. If you are short on time, read only the 🏛️ rows (there are about a dozen) and you have the spine of the year.

🏛️ = landmark / load-bearing 🛠️ = method / recipe 📦 = model report 🔬 = analysis / study
Q1 · W1–16 MoE wave + GRPO + 1-bit + long ctx Q2 · W17–27 architecture bets Mamba-2, KAN, MLA Q3 · W28–37 model-report summer Llama 3, test-time Q4 · W38–52 reasoning + efficiency o1 era, V3, latent CoT

Fig 1 — The four moods of 2024, one quarter at a time.

Q1 · January – March (weeks 1–16)

The year opens with the mixture-of-experts wave and, quietly, the single most consequential method of the decade: GRPO, buried inside a math paper. Also the 1-bit experiments and the first 2M-token context tricks.

WkPaperWhy read
W1 · Jan 📦Mixtral of ExpertsThe open MoE that made sparse models mainstream — 8 experts, 2 active. The template every 2024 MoE follows.
W2 · Jan 🛠️DeepSeekMoEFine-grained experts + shared experts for common knowledge. The MoE recipe DeepSeek rode all the way to V3.
W3 · Jan 🛠️Self-Play Fine-Tuning (SPIN)A model improves by playing against its own past outputs — no new human data. Early seed of self-improvement.
W4 · Jan 🛠️Self-Rewarding Language ModelsThe model is its own judge, generating its own rewards during iterative DPO. The idea behind a whole 2025 sub-field.
W5 · Jan 🔬RAG vs Fine-TuningThe clean comparison everyone cites when deciding retrieval vs weights. Settles a lot of arguments.
W6 · Feb 🏛️DeepSeekMath (introduces GRPO)The most important 2024 paper for what came next — GRPO, the RL algorithm behind DeepSeek-R1 and the entire 2025 reasoning boom. Read this one closely.
W7 · Feb 🏛️The Era of 1-bit LLMs (BitNet b1.58)Ternary {-1,0,1} weights match fp16 at the same size. A genuinely radical efficiency claim that still echoes in 2026.
W8 · Feb 🛠️Self-DiscoverLLMs compose their own reasoning structures from atomic modules. Pre-R1 reasoning, prompt-only.
W9 · Feb 🛠️DoRA: Weight-Decomposed LoRASplits weight magnitude and direction for better PEFT — a near-free upgrade over plain LoRA.
W10 · Feb 🛠️LongRoPEExtends context past 2M tokens by reworking positional interpolation. The long-context groundwork.
W11 · Feb 🛠️OneBit: Extremely Low-bit LLMsPushes weight quantization to ~1 bit. Read alongside BitNet to see the low-bit frontier.
W12 · Feb 🏛️Genie: Generative Interactive EnvironmentsAn 11B world model that turns videos into playable worlds — no action labels. The foundation-world-model opener.
W13 · Mar 📦Gemini 1.51M-token multimodal context. The release that made "long context" a headline capability.
W14 · Mar 🛠️GaLoreFull-parameter training in a fraction of the memory via low-rank gradients. Train bigger on less.
W15 · Mar 🛠️Vision-RWKVAn RNN-style linear-attention backbone for vision — early signal of the post-transformer architecture interest.
W16 · Mar 🏛️Stable Diffusion 3 (Rectified Flow Transformers)The MM-DiT architecture + rectified flow that set the text-to-image standard. The image-gen landmark of the year.

Q2 · April – June (weeks 17–27)

The architecture-bet quarter. Mamba-2 unifies SSMs with attention, KAN questions the MLP itself, DeepSeek-V2 ships Multi-head Latent Attention, and the alternatives (xLSTM, Mixture-of-Depths) pile up.

WkPaperWhy read
W17 · Apr 🏛️KAN: Kolmogorov-Arnold NetworksLearnable activations on edges instead of fixed ones on nodes — a from-scratch rethink of the MLP. The year's most-discussed "what if everything is different" paper.
W18 · Apr 🛠️Infini-attentionUnbounded context with bounded memory via compressive memory. The "infinite context" idea, distilled.
W19 · Apr 🛠️Mixture-of-DepthsTokens dynamically skip layers under a compute budget. MoE's lesser-known cousin — compute where it's needed.
W20 · Apr 📦Phi-3 Technical ReportA 3.8B model rivaling Mixtral/GPT-3.5, runnable on a phone. The data-quality-over-scale thesis, demonstrated.
W21 · May 🏛️📦DeepSeek-V2 (MLA)Multi-head Latent Attention — ~93% KV-cache reduction. The attention innovation the whole DeepSeek line is built on.
W22 · May 🏛️Mamba-2 (Transformers are SSMs)State-space duality unifies attention and SSMs; the core layer runs 2–8× faster. The theoretical bridge that 2026's Mamba revival stands on.
W23 · May 🛠️xLSTM: Extended LSTMScales the LSTM to billions of params with exponential gating. The "what if RNNs were never the problem" bet.
W24 · May 🛠️SimPOReference-free preference optimization — simpler and often stronger than DPO. The practical alignment upgrade of the year.
W25 · May 📦Chameleon: Mixed-Modal Early-FusionOne model, text and images interleaved from the start (not bolted on). The early-fusion multimodal direction.
W26 · Jun 📦Nemotron-4 340BNVIDIA's open 340B family — and a synthetic-data masterclass (98%+ of alignment data is generated). Read for the data pipeline.
W27 · Jun 🛠️RouteLLMLearn to route each query to a strong or weak model — 2× cost cuts with no quality loss. The pragmatic cost play.

Q3 · July – September (weeks 28–37)

The model-report summer — Llama 3, Qwen2, Gemma 2 — plus the paper that reframed the whole roadmap: scaling test-time compute can beat scaling parameters.

WkPaperWhy read
W28 · Jul 🏛️📦The Llama 3 Herd of ModelsMeta's 90-page report on building a frontier open model end to end. The most complete public model report of the year.
W29 · Jul 📦Qwen2The release that began Qwen's rise to the dominant open family. Baseline for everything Qwen ships after.
W30 · Aug 📦Gemma 2Google's open report — distillation + local/global attention. The small-open reference before Gemma 3.
W31 · Aug 🏛️Scaling LLM Test-Time Compute OptimallySpending compute at inference can beat scaling parameters by 4×+. The paper that made test-time scaling a first-class axis — and set up o1.
W32 · Aug 📦Jamba-1.5 (Hybrid Transformer-Mamba)A production hybrid SSM-transformer at scale. The blueprint 2026's hybrid models (Nemotron 3) follow.
W33 · Aug 🏛️The AI ScientistSakana's end-to-end automated research loop — idea → code → experiment → paper → review. The "AI does science" milestone (precursor to AlphaEvolve).
W34 · Sep 📦OLMoE: Open MoEFully open MoE — weights, data, code, logs. The most transparent MoE to learn the architecture from.
W35 · Sep 📦Qwen2-VLDynamic resolution + M-RoPE for images and video. The open VLM reference of late 2024.
W36 · Sep 🛠️SCoRe: Self-Correct via RLDeepMind teaches genuine self-correction with multi-turn RL on self-generated data. The serious version of "think again."
W37 · Sep 🔬In Defense of RAG in the Long-Context EraEven with million-token windows, retrieval still wins on cost and accuracy. Settles the "is RAG dead?" debate.

Q4 · October – December (weeks 38–52)

The finale, and the densest stretch. o1 lands and the reasoning race begins; Apple punctures the hype with GSM-Symbolic; and December alone ships DeepSeek-V3, latent-space reasoning, byte-level models, and a modern BERT. This quarter is the launchpad for everything in the 2025 list.

WkPaperWhy read
W38 · Oct 🏛️Differential TransformerAttention as the difference of two softmaxes — cancels noise, sharpens focus, cuts hallucination. A clean, adoptable architecture tweak.
W39 · Oct 🔬GSM-Symbolic (Apple)Shows models wobble when you just rename variables — questioning whether GSM8K gains are real reasoning. The skeptic anchor for the o1 era.
W40 · Oct 🛠️O1 Replication JourneyAn open, honest attempt to reverse-engineer o1-style reasoning. The bridge from o1's announcement to R1's open recipe.
W41 · Nov 🏛️🛠️Tülu 3A fully open post-training recipe — SFT, DPO, and the first big use of RL with Verifiable Rewards (RLVR). The open playbook the 2025 RLVR wave standardized.
W42 · Nov 🛠️BitNet a4.84-bit activations for 1-bit models — pushing the low-bit frontier toward real deployment. Pairs with W7.
W43 · Nov 🛠️Marco-o1An open o1-style reasoner with MCTS + reflection, aimed at open-ended (not just verifiable) tasks. Early open-reasoning attempt.
W44 · Dec 🏛️📦DeepSeek-V3671B MoE (37B active), MLA + auxiliary-loss-free load balancing + multi-token prediction, trained shockingly cheap. The base model R1 was built on. Essential.
W45 · Dec 🏛️Coconut: Reasoning in Continuous Latent SpaceReason by feeding hidden states back in, not tokens — thinking without words. A genuinely new reasoning paradigm.
W46 · Dec 🏛️Byte Latent Transformer (BLT)Tokenizer-free, entropy-based byte patches that match BPE at scale. The strongest case yet for killing the tokenizer.
W47 · Dec 🏛️ModernBERTThe encoder finally modernized — RoPE, local/global attention, 8K context, 2T tokens. The new default for retrieval/classification.
W48 · Dec 📦Qwen2.5 Technical Report18T-token pretraining, 1M+ SFT samples, multistage RL. The report that set up Qwen3. The most-downloaded open family's blueprint.
W49 · Dec 📦Phi-4A 14B model that surpasses its GPT-4 teacher on STEM via synthetic-data curriculum. The data-centric thesis at its peak.
W50 · Dec 🛠️Large Concept Models (Meta)Predict sentence-level concepts in an embedding space instead of tokens. An ambitious bet on a higher unit of language.
W51 · Dec 🛠️Deliberative Alignment (OpenAI)Teach the model to reason over the safety spec before answering — reasoning as a safety tool. How the o-series is aligned.
W52 · Dec 📦DeepSeek-VL2DeepSeek's MoE vision-language model — closes out the year and the multimodal thread. A fitting week 52.

The load-bearing dozen

If 52 is too many, read only these 🏛️ papers. They are the ones 2025 and 2026 cite on every page — the actual spine of the year.

#PaperWhat it set up
1DeepSeekMath / GRPOThe RL algorithm behind R1 and the entire 2025 reasoning era.
2DeepSeek-V2 / MLAThe cheap-attention innovation the DeepSeek line rides.
3DeepSeek-V3The base model R1 was fine-tuned from.
4Scaling Test-Time ComputeMade inference compute a scaling axis — the o1 thesis.
5Mamba-2The SSM/attention bridge behind 2026's Mamba revival.
6BitNet b1.58The 1-bit efficiency frontier still cited in 2026.
7Tülu 3Open post-training + the first big RLVR demonstration.
8CoconutLatent-space reasoning — a new paradigm 2025 builds on.
9Llama 3 HerdThe reference for how to build a frontier open model.
10GSM-SymbolicThe reasoning-skepticism anchor for the o1 era.
11Byte Latent TransformerThe strongest case for tokenizer-free models.
12KANThe year's boldest "rethink the building block" paper.

Takeaways

2024 built the launchpad; 2025 lit the fuse. Nearly every big 2025 story traces to a 2024 paper here — GRPO (W6) became R1, test-time compute (W31) became o1 and the reasoning race, DeepSeek-V3 (W44) became R1's base, Tülu 3's RLVR (W41) became the standard signal, and the 2024 architecture bets (Mamba-2, BitNet, MLA) became the 2026 efficiency era.

Read it as a calendar, not a pile. One paper a week. By December you will have watched MoE go mainstream, the tokenizer come under threat, reasoning get an algorithm, and efficiency become the headline — which is exactly the run-up the 2025–2026 list continues.

References & extra reads

A curated 52 of 2024, one per week. Dates are first arXiv submission month; links go to abstract pages so you get versions, code, and BibTeX. Emoji tags (🏛️🛠️📦🔬) are my own triage marks. Coverage informed by the public 2024 roundups, then cut and re-picked for impact.

← prev: the 2025-2026 reading list next: Terraform 1.15, in depth →
© cvam — written in plaintext, served warm