Skip to content

Benchmark

BABILong qa4 64k

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

Models and systems measured
10
Scores recorded
10
Sources
1
IQ weight
No weight published

Score distribution

54.5%
Mean
51%
Median
25.5%
Std dev
39%
10th percentile
100%
90th percentile
13%
Min
13%100%
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-tuneqa4, 64k· BABILong v0 data docs100%
~ Mamba (130M) fine-tuneqa4, 64k· BABILong v0 data docs100%
~ RMT (137M) fine-tuneqa4, 64k· BABILong v0 data docs51%
ChatGLM3 6Bchatglm3-6b-128k, qa4, 64k, 128k· BABILong v0 data docs13%
GPT-4-0125-previewqa4, 64k· BABILong v0 data docs40%
Llama3-ChatQA-1.5-8B + RAGqa4, 64k· BABILong v0 data docs59%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B-Instruct-v0.1, qa4, 64k· BABILong v0 data docs39%
Phi-3 Medium 128K InstructPhi-3-medium-128k-instruct, qa4, 64k· BABILong v0 data docs52%
Phi-3 Mini 128K InstructPhi-3-mini-128k-instruct, qa4, 64k· BABILong v0 data docs47%
Yi-34B01-ai/Yi-34B-200k, qa4, 64k, 200k· BABILong v0 data docs44%

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