Unverified50% confidenceBenchmarkExact time
With no_repeat_ngram_size: 3 enabled alone, Llama 3.2 1B's JSON pass rate was 1/24 (4%) and SmolLM2 360M's was 0/24 (0%), making it the main cause of the JSON output collapse.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe R9700 achieved a median decoding speed of 153.1 tok/s on JSON tasks.62% similarUnverifiedDuring the LLM sequential decoding stage, a pipeline-parallel cluster exchanges activation tensors rather than model weights; taking hidden dimension d=3584 as an example, at FP32 precision this is about 14.3KB per token, and after symmetric INT8 dynamic quantization it is only 5.1KB per hop on the network.59% similarUnverifiedNVFP4 is actually approximately 3.75 bits, and is not strictly a 4-bit quantization in the rigorous sense.55% similarUnverifiedThe arithmetic intensity of the decoding stage in LLM autoregressive generation is extremely low: for a 7B-parameter FP16 model, one forward pass requires loading approximately 14 GB of weights but executes only about 14 GFLOPs of computation, resulting in an arithmetic intensity of approximately 1 FLOP/Byte, making it a bandwidth-bound workload.55% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1056264API
curl https://kongchang.com/api/v1/knowledge/claims/1056264MCP
get_claim(id=1056264)