List your company

Chart Question Answering on ChartQA (Accuracy)

90.49AccuracyMetaForge
63.834870.754977.67584.5951Sep 28, 2025Nov 13, 2025Dec 29, 2025Feb 14, 2026Apr 1, 2026May 17, 2026Jul 3, 2026

Evaluation results

MethodLinks
MetaForgeTool Setting=w/ IID Tools2026.0690.49
MetaForgeTool Setting=w/ OOD Tools2026.0690.18
BF16Model=Kimi-VL-A3B-Instruct, Avg-Bit=BF162026.0689.44
SigLIP2-gTokens per frame/tile=256, TTFT (ms)=517.6, LLM Backbone=9B Nemotron2026.0789.4
RADIO1DTokens per frame/tile=192, TTFT (ms)=380.4, LLM Backbone=9B Nemotron2026.0789.2
RADIO1DTokens per frame/tile=224, TTFT (ms)=410.2, LLM Backbone=9B Nemotron2026.0789.2
C-RADIOv4-HTokens per frame/tile=256, TTFT (ms)=468.2, LLM Backbone=9B Nemotron2026.0789.1
RADIO1DTokens per frame/tile=256, TTFT (ms)=452.8, LLM Backbone=9B Nemotron2026.0789.1
SigLIP2-SO400mTokens per frame/tile=256, TTFT (ms)=440.5, LLM Backbone=9B Nemotron2026.0788.8
RADIO1DTokens per frame/tile=128, TTFT (ms)=373.2, LLM Backbone=9B Nemotron2026.0788.6
RADIO1DTokens per frame/tile=64, TTFT (ms)=346.1, LLM Backbone=9B Nemotron2026.0787.9
FP16 (Baseline)Backbone=Qwen 2.5 VL 7B, Avg. Bit width=162025.0987.3
MODEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0687.26
InternVL3Size=8B2026.0686.6
LLaVA-OneVision-1.5Size=8B2026.0686.5
GPT-5-highSize=-2026.0686.2
RADIO1DTokens per frame/tile=32, TTFT (ms)=335.1, LLM Backbone=9B Nemotron2026.0785.8
VEQ-MEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0685.64
BF16Model=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=BF162026.0685.2
AWQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=42025.0984.5
MC-MoEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0684.26
GPTQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=42025.0984.1
MoEQuantModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0683.77
Claude-Opus-4.1Size=-2026.0683.3
MBQModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0683.13
GreedyModel=Qwen2.5-VL (3B)2026.0483.12
MODEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0683
Speed-QModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0682.97
OmniQuantBackbone=Qwen 2.5 VL 7B, Avg. Bit width=42025.0982.9
RADIO1DTokens per frame/tile=8, TTFT (ms)=329.7, LLM Backbone=9B Nemotron2026.0782.8
GreedyBackbone=Qwen2.5-VL (7B)2026.0482.13
GPTQModel=Kimi-VL-A3B-Instruct, Avg-Bit=W32026.0682.04
Top-p (nucleus)Model=Qwen2.5-VL (3B), tau=0.72026.0481.13
VEQ-MEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0681.12
ε-SamplingTemperature (τ)=0.7, Backbone=Qwen2.5-VL (7B)2026.0480.16
η-SamplingTemperature (τ)=0.7, Backbone=Qwen2.5-VL (7B)2026.0480.16
Top-p (nucleus)Temperature (τ)=0.7, Backbone=Qwen2.5-VL (7B)2026.0480
MC-MoEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0679.94
ε-SamplingModel=Qwen2.5-VL (3B), tau=0.72026.0479.91
Top-p (nucleus)Temperature (τ)=1.0, Backbone=Qwen2.5-VL (7B)2026.0479.69
η-SamplingModel=Qwen2.5-VL (3B), tau=0.72026.0479.66
Min-pModel=Qwen2.5-VL (3B), tau=0.72026.0479.62
Top-kModel=Qwen2.5-VL (3B), tau=0.72026.0479.61
Temp' OnlyModel=Qwen2.5-VL (3B), tau=0.72026.0479.58
Temp’ OnlyTemperature (τ)=0.7, Backbone=Qwen2.5-VL (7B)2026.0478.96
Top-kTemperature (τ)=0.7, Backbone=Qwen2.5-VL (7B)2026.0478.96
Top-p (nucleus)Model=Qwen2.5-VL (3B), tau=1.02026.0478.93
η-SamplingTemperature (τ)=1.0, Backbone=Qwen2.5-VL (7B)2026.0478.73
SMMDBackbone=Qwen2.5-VL, Parameters=7B2026.0678.38
ε-SamplingTemperature (τ)=1.0, Backbone=Qwen2.5-VL (7B)2026.0478.18
Temp’ OnlyTemperature (τ)=1.0, Backbone=Qwen2.5-VL (7B)2026.0477.97
Top-kTemperature (τ)=1.0, Backbone=Qwen2.5-VL (7B)2026.0477.92
NTILBackbone=Qwen2.5-VL, Parameters=7B2026.0677.08
CEBackbone=Qwen2.5-VL, Parameters=7B2026.0677.02
NTLBackbone=Qwen2.5-VL, Parameters=7B2026.0676.92
ε-SamplingModel=Qwen2.5-VL (3B), tau=1.02026.0476.53
η-SamplingModel=Qwen2.5-VL (3B), tau=1.02026.0476.46
Temp' OnlyModel=Qwen2.5-VL (3B), tau=1.02026.0476.14
Top-kModel=Qwen2.5-VL (3B), tau=1.02026.0476.11
Min-pModel=Qwen2.5-VL (3B), tau=1.02026.0476.09
GCEBackbone=Qwen2.5-VL, Parameters=7B2026.0675.87
MODEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W22026.0675.2
CoLTSize=8B, Backbone=Qwen3-VL-8B2026.0674.7
SMMDBackbone=Ministral-3, Parameters=3B2026.0673.21
Speed-QModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0673.2
MODEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W22026.0673.11
NTILBackbone=Qwen2.5-VL, Parameters=3B2026.0672.95
GCEBackbone=Ministral-3, Parameters=3B2026.0672.32
SMMDBackbone=Qwen2.5-VL, Parameters=3B2026.0672.21
NTLBackbone=Qwen2.5-VL, Parameters=3B2026.0672.01
NTILBackbone=Ministral-3, Parameters=3B2026.0671.95
CEBackbone=Ministral-3, Parameters=3B2026.0671.64
Min-pTemperature (τ)=1.0, Backbone=Qwen2.5-VL (7B)2026.0471.56
Min-pTemperature (τ)=2.0, Backbone=Qwen2.5-VL (7B)2026.0471.51
CEBackbone=Qwen2.5-VL, Parameters=3B2026.0671.44
Min-pTemperature (τ)=0.7, Backbone=Qwen2.5-VL (7B)2026.0471.41
MoEQuantModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0671.4
GCEBackbone=Qwen2.5-VL, Parameters=3B2026.0670.44
Qwen2.5-VL-3Bvariant=MoDoMoDo2025.1070.4
RECAPbackbone=Qwen2.5-VL-3B2025.1070.4
Qwen2.5-VL-3B-Instruct-GRPO-deepmath2025.1070.24
ε-SamplingTemperature (τ)=2.0, Backbone=Qwen2.5-VL (7B)2026.0469.79
Qwen2.5-VL-3Bvariant=Uniform2025.1069.68
NTLBackbone=Ministral-3, Parameters=3B2026.0669.66
LUQBackbone=Qwen 2.5 VL 7B, Avg. Bit width=2.75, Quantized Layers=12 layer2025.0968.6
RADIO1DTokens per frame/tile=1, TTFT (ms)=327.0, LLM Backbone=9B Nemotron2026.0768.2
MBQModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0668.16
vision-grpo-qwen-2.5-vl-3b2025.1067.8
Top-p (nucleus)Temperature (τ)=2.0, Backbone=Qwen2.5-VL (7B)2026.0467.6
LVRSize=7B2026.0667.2
VEQ-MEModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W22026.0667.16
SIM-CoTSize=8B, Backbone=Qwen3-VL-8B2026.0666.7
MC-MoEModel=Kimi-VL-A3B-Instruct, Avg-Bit=W22026.0666.49
η-SamplingTemperature (τ)=2.0, Backbone=Qwen2.5-VL (7B)2026.0466.2
LaViDa-O + RL (Global + Factorized)Arch=D-Diff2026.0666.16
GPTQModel=Qwen3-VL-30B-A3B-Instruct, Avg-Bit=W32026.0666.16
Qwen3-VL (Textual reasoning)Size=8B2026.0665.1
GCEBackbone=Llama-3.2-Vision, Parameters=11B2026.0664.91
SMMDBackbone=Llama-3.2-Vision, Parameters=11B2026.0664.91
CEBackbone=Llama-3.2-Vision, Parameters=11B2026.0664.86