supersquirrel

cross-posted from: https://lemmy.ca/post/69219331

Among other things github interaction with projects is analyzed so I think this is relevant to programming too.

open access paper https://arxiv.org/abs/2511.03877

Cross-channel prediction outperforms same-channel pre- diction for early input-horizon, across all models. This is consistent with correlations plots in Figure 3 and Figure 2....

...

We establish Lead-Lag Forecasting (LLF) as a formal prediction problem, motivated by the gap between observed lead-lag dynam- ics in important domains—including scientific and technological impact—and popular time series forecasting benchmarks. We cat- alyze research on LLF by curating and releasing two novel datasets: arXiv papers and GitHub repositories. We establish lead-lag rela- tionships in streams of activity data and provide baseline numbers for several standard supervised machine learning methods on the task of predicting a 5-year outcome from as little as one month of observation. While our results demonstrate the existence of predic- tive signal, we speculate that there are opportunities for innovation to improve predictions.

Smells like Category Theory to me!

"It's very hard to collect data that trains the model to be better than humans, because there are very few humans who can create that data," Raj said. "You want to make it better than a Fields Medalist or a Nobel Prize winner. How do you collect that?"

...

While there, he worked alongside other researchers to understand the failure points of models hosted on cloud computing infrastructure and offer specific solutions, he said. He also created "synthetic" data that, unlike human-created writing or code, is generated artificially before being reused as training material for the LLMs, Raj said.

Creating synthetic data that matches the quality of human-created data is a tall task, Raj said. Computing power can be acquired relatively easily, but finding the data to support the project is uncharted territory, he added.

sigh

Garbage in garbage out, even if the garbage is synthetic that doesn't make it not garbage...?

open access paper https://arxiv.org/abs/2511.03877

Cross-channel prediction outperforms same-channel pre- diction for early input-horizon, across all models. This is consistent with correlations plots in Figure 3 and Figure 2....

...

We establish Lead-Lag Forecasting (LLF) as a formal prediction problem, motivated by the gap between observed lead-lag dynam- ics in important domains—including scientific and technological impact—and popular time series forecasting benchmarks. We cat- alyze research on LLF by curating and releasing two novel datasets: arXiv papers and GitHub repositories. We establish lead-lag rela- tionships in streams of activity data and provide baseline numbers for several standard supervised machine learning methods on the task of predicting a 5-year outcome from as little as one month of observation. While our results demonstrate the existence of predic- tive signal, we speculate that there are opportunities for innovation to improve predictions.

Smells like Category Theory to me!