TAG COLLECTION

long context

All AI Pulse updates tagged "long context".

4 signals

X.com10/15, 12:51Model Release

Ling-3.1-flash open sourced with 1M context

Ling-3.1-flash is open source with about 560B total parameters, 25B active per token and up to 1M context, aimed at turning briefs into pages, mobile UI or prototypes.

Signal score 100Source

X.com09/29, 22:31Research

Long-context models struggle to find corpus contradictions

sewon__min says long-context LMs do well on RULER-like tasks but suffer at finding every contradiction as corpora scale, even with in-domain fine-tuning.

long context RULER evaluation language models

Signal score 100Source

X.com09/23, 20:19Model Release

MiMo-V3 adopts HySparse2 for cheaper long context

MiMo-V3 is getting a new architecture: core HySparse2 cuts prefill and KV cache and improves long-context retrieval versus MiMo-V2.6 Hybrid SWA, with 5.02× lower prefill FLOPs at 1M tokens.

MiMo-V3 HySparse2 KV cache long context

Signal score 100Source

X.com06/12, 23:48Infrastructure

AgentPerf treats agentic AI as a workload

AgentPerf treats agentic AI as a long-running workload, considering memory and networking.

rohanpaul_ai AI MoE long context

Signal score 13Source