day | topic | resources | details |
Apr 25 | introduction | organization, lectures, student topics | |
May 2 | synthetic data | talk by Latif Köksal, DeepMind | |
assignment of topics | |||
May 9 | memory etc. | NoLiMa | talk by Ali Modaresi |
May 16 | multilinguality (1) | Manchu | talk by Peiqin Lin |
May 23 | multilinguality (2) | representation&editing | talk by Mingyang Wang |
May 30 | |||
Jun 6 | |||
Jun 13 | |||
Jun 20 | |||
Jun 27 | |||
Jul 4 | |||
Jul 11 | |||
Jul 18 | |||
Jul 25 |
paper | topic | ||
all topics covered in the lectures (see above) | |||
Geiping et al. | test-time compute: recurrent depth approach | ||
DeepSeek-AI | DeepSeek-R1: reasoning through RL | ||
Gemma Team | Gemma 3 technical report | ||
Olsson et al. | induction heads | ||
Maini et al. | rephrasing the web | ||
Sharkey et al. | open problems in mechanistic interpretability | ||
Park et al. | linear representation hypothesis | ||
Han et al. (2024) | word embeddings are steers | ||
Llama Team (2024) | Llama 3 | ||
Qwen et al. (2024) | Qwen 2.5 | ||
Abdin et al. (2024) | Phi-4 | ||
Makelov et al. (2024) | sparse autoencoders (2) | ||
McDougall et al. (2023) | copy suppression | ||
Saphra et al. (2024) | notion of mechinterp | ||
Dutta et al. (2024) | mechinterp: COT | ||
Geva et al. (2023) | factual associations/enrichment | ||
nostalgebrist (2020) | logit lens | ||
Chughtai et al. (2024) | summing up the facts | ||
Shao et al. (2024) | DeepSeekMath | ||
Zhao et al. (2024) | Marco-o1 | ||
Wu et al. (2024) | REFT | ||
Hübotter et al. (2025) | SIFT | ||
Hughes et al. (2024) | open-endedness | ||
Turpin et al. (2023) | unfaithful COT | ||
Gottweis et al. (2025) | AI co-scientist | ||
Venhoff et al. (2025) | steered reasoning | ||
Yu et al. (2024) | superweights | ||
Bricken et al. (2023) | sparse autoencoders (1) | ||
Packer et al. (2023) | MemGPT | ||
Milliere et al. (2024) | philosophy of LLMs | ||
StanfordNLP (2024) | DSPy | ||
Durrani et al. (2020) | analyzing neurons | ||
Voita et al. (2023) | dead neurons | ||
De Peuter et al. (2023) | human-agent cooperation | ||