Writing

Plain-language write-ups of the lab's papers and methods, published on members' own sites. For the papers themselves, see Publications.

How to Create a Simple Synthetic Dataset How the LAPSE benchmark was built: small probes first, minimal pairs, checked invented names, and a smoke test read by hand. Sugam Panthi · [READ] Every Reported Number Should Trace to One Scorer and Run The bookkeeping rules behind running several paper projects at once, so every number in a paper points to the run that produced it. Sugam Panthi · [READ] LLMs Absorb Text Typed After a Paste Why models edit a typed afterthought into a pasted document, and why explicit markers work where blank lines do not. Sugam Panthi · [READ] Recovery Tools Help Agents Respond to Corrupted Tool Results Agents trust plausible but wrong tool output; a monitor helps only when its receipt names something the agent can do next. Sugam Panthi · [READ] Memory Consolidation Turns Temporary Statements Into Standing Facts A LAPSE pilot: memory writers rewrite "I am staying in" as "lives in", and date stamps barely change it. Sugam Panthi · [READ] Scoring Targets Change Which Memory System Wins The same ranked list can crown a different winner when the scoring target changes. Sugam Panthi · [READ] Evidence Compression Is Reader-Dependent Compressing retrieved evidence helps small reader models and is a coin flip for strong ones. Sugam Panthi · [READ]