Perceiver IO: A General Architecture for Structured Inputs & Outputs · HackerPapers discussions

Perceiver IO: A General Architecture for Structured Inputs & Outputs — HN submission

Wed, 02 Apr 2025 23:26:52 GMT

Perceiver IO: A General Architecture for Structured Inputs and Outputs (2021)

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Perceiver IO: A General Architecture for Structured Inputs & Outputs — mentioned by tehsauce

Sat, 03 Jun 2023 16:45:56 GMT

Seems in similar spirit to the “perceiver” architecture from deepmind a couple years ago: https://arxiv.org/abs/2107.14795

Paper and discussion history. A mention is not an endorsement.

Perceiver IO: A General Architecture for Structured Inputs & Outputs — mentioned by gwern

Sun, 28 Nov 2021 18:04:17 GMT

I think if you wanted to do this at scale, you would prefer Perceiver-IO ( https://arxiv.org/abs/2103.03206#deepmind https://arxiv.org/abs/2107.14795#deepmind ).

Paper and discussion history. A mention is not an endorsement.

Perceiver IO: A General Architecture for Structured Inputs & Outputs — mentioned by jeeceebees

Mon, 13 Sep 2021 08:03:03 GMT

There is a lot of evidence that these token-based models work with multi-modal data. In fact, several groups have proposed different multi-modal transformer architectures already (e.g. [1] or [2]), although I don't believe anyone has scaled them up much farther than 300M parameters yet. If these models are shown videos of butterflies flapping their wings with a text description of 'a butterfly flapping its wings,' why wouldn't you expect it to start to relate the information coming from multiple modalities? It's definitely a challenge to get enough high-quality data to feed a 100B parameter…

Paper and discussion history. A mention is not an endorsement.

Perceiver IO: A General Architecture for Structured Inputs & Outputs — HN submission

Mon, 02 Aug 2021 02:08:02 GMT

Perceiver IO: A General Architecture for Structured Inputs and Outputs

2 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.