alem-world/alem-llm-traces

Alem — LLM Reasoning Traces Full per-step interactive debug viewers for language agents evaluated on Alem, a JAX benchmark for open-ended multi-agent coordination. Each file is a self-contained HTML page that replays one episode step by step, showing, for all three agents at every turn: the rendered frame (world view), the observation given to the model (long- and short-term), the model's reasoning, chosen action, communication broadcast, and scratchpad memory, the messages… See the full description on the dataset page: https://huggingface.co/datasets/alem-world/alem-llm-traces.

<iframe src="https://silytics.si/embed/item/38803a36-4684-47e1-9e55-6f7e47a0b946" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.