Hypernym (@HypernymAI) profile picture

We build breakthroughs in long-context for infinite memory + advanced inference solving model scale, context depth & lost in the middle.

1 AU
Joined August 2024
Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)Photo by Hypernym (@HypernymAI)
The common belief today is that once models exceed 256k tokens, context performance falls off a cliff. At Hypernym, we’ve internally proven this widely accepted “fact” is false. It’s not a technical limitation — it’s simply the result of an industry moving too fast. No matter which model we apply our infrastructure to, LITM drastically improves — among many other things. More to come. 🚀
2
9
455
🎧 New research drop: Double feature Side A: LLMs have fingerprints. Side B: Agents hate poetry. This project shows that you can tell which model is responding, even if it’s been quietly swapped. Wait, there's more: when prompts get too poetic? Tool-using agents degrade. 📄 Built by Corpaci Luiza Cristina, Siddhesh Pawar, & @Elationate ⚡️ Powered by @HypernymAI 🔬 Hosted by @apartresearch and @withmartian 👉 Read: apartresearch.com/project/ll…
1
7
176