Skip to content

Benchmark

BABILong qa3 8k

Long context benchmarks measure how well models use information in long inputs. Scores are percentages. Higher is better.

Models and systems measured
14
Scores recorded
15
Sources
1
IQ weight
No weight published

Score distribution

38.3%
Mean
27%
Median
26.5%
Std dev
16%
10th percentile
92%
90th percentile
11%
Min
11%97%
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
~ ARMT (137M) fine-tuneqa3, 8k· BABILong v0 data docs92%
~ Mamba (130M) fine-tuneqa3, 8k· BABILong v0 data docs97%
~ RMT (137M) fine-tuneqa3, 8k· BABILong v0 data docs73%
ChatGLM3 6Bchatglm3-6b-128k, qa3, 8k, 128k· BABILong v0 data docs27%
GPT-4-0125-previewqa3, 8k· BABILong v0 data docs53%
Llama 2 7bLLaMA-2-7B-32K, qa3, 8k, 32k· BABILong v0 data docs27%
Llama 3 8B InstructMeta-Llama-3-8B-Instruct, qa3, 8k· BABILong v0 data docs11%
Llama3-ChatQA-1.5-8B + RAGqa3, 8k· BABILong v0 data docs16%
Mistral 7B Instruct v0.2Mistral-7b-Instruct-v0.2, qa3, 8k· BABILong v0 data docs24%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa3, 8k· BABILong v0 data docs31%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B-Instruct-v0.1, qa3, 8k· BABILong v0 data docs27%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa3, 8k· BABILong v0 data docs30%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa3, 8k· BABILong v0 data docs26%
Yi-34B01-ai/Yi-34B-200k, qa3, 8k, 200k· BABILong v0 data docs24%
Yi-34B01-ai/Yi-34B, qa3, 8k· BABILong v0 data docs17%

Data credits: RMT-team, Kuratov et al. Official v0 numeric CSV; results licence not separately stated.