X-VLA authors
X-VLA
What X-VLA’s own documents say it is built from — every method with the sentence that describes it.
- Parameters
- 0.9B
- Attention
- Florence-2-large backbone; soft prompts per embodiment
- Input
- text+image -> action
- Licence
- Apache-2.0
- Released
- 2025-11-04
Against the consensus recipe
Its documents state 0 of the 18 methods the field agrees on.
Data curation
- Benchmark decontaminationnot in its documents
- Synthetic data generationnot in its documents
- Corpus deduplicationnot in its documents
Training objective
- Multi-Token Predictionnot in its documents
- Knowledge insulationnot in its documents
- Quantization-aware distillationnot in its documents
Model architecture
- Mixture of expertsnot in its documents
- Multi-Token Predictionnot in its documents
- Grouped-query attentionnot in its documents
Optimization
- Muonnot in its documents
- Expert parallelismnot in its documents
- Quantization-Aware Trainingnot in its documents
Post-training
- Supervised fine-tuningnot in its documents
- Asynchronous reinforcement learningnot in its documents
- Multi-Teacher On-Policy Distillationnot in its documents
Inference & serving
- Thinking modenot in its documents
- Multi-Token Predictionnot in its documents
- FP8not in its documents
Checked in its code
Its code shows 0 of the 19 architecture features checked.
Sources say what a lab chose to describe; the modeling code says what the checkpoint runs. Each feature below was put to a code verifier, and only the config values and source lines it cited that were found in the files are shown, linked to the line.
Read from the model repository’s own code (modeling_xvla.py, modeling_florence2.py, configuration_florence2.py, transformer.py, action_hub.py, configuration_xvla.py) and config.json at revision c1c4a64a7e, by anthropic/claude-sonnet-5 on 2026-11-07.
| Feature | Code | Its documents | Verdict |
|---|---|---|---|
| Grouped-query attention | not in its code | not stated | The Florence2 text encoder/decoder attention and the SoftPromptedTransformer attention both use standard multi-head attention with equal… show evidenceThe Florence2 text encoder/decoder attention and the SoftPromptedTransformer attention both use standard multi-head attention with equal numbers of query, key and value heads, with no separate kv head count. |
| Multi-head latent attention | not in its code | not stated | No low-rank KV latent projection exists anywhere in the attention implementations; keys/values are projected directly to full embed_dim as… show the whole verdictNo low-rank KV latent projection exists anywhere in the attention implementations; keys/values are projected directly to full embed_dim as in standard attention. |
| Sliding-window attention | not in its code | not stated | No sliding-window masking mechanism is implemented in the Florence2 text attention or the action transformer attention; attention is… show the whole verdictNo sliding-window masking mechanism is implemented in the Florence2 text attention or the action transformer attention; attention is either full bidirectional (encoder) or full causal (decoder). |
| Interleaved sliding-window and global attention | not in its code | not stated | There is no interleaving of sliding and full-attention layers; all layers use the same uniform full/causal attention pattern. |
| Indexer-selected sparse attention (DeepSeek Sparse Attention) | not in its code | not stated | No indexer-based top-k token selection mechanism is present in any attention module. |
| Linear-attention or state-space layers alongside full attention | not in its code | not stated | All token-mixing layers (Florence2 encoder/decoder, action transformer) use standard softmax attention; no linear-attention or state-space… show the whole verdictAll token-mixing layers (Florence2 encoder/decoder, action transformer) use standard softmax attention; no linear-attention or state-space mixers are implemented. |
| Gated DeltaNet layers | not in its code | not stated | No gated delta-rule recurrent mixer is implemented anywhere in the provided files. |
| Mamba-2 layers | not in its code | not stated | No Mamba-2 or SSM-style selective state-space block is implemented in the given modeling files. |
| Learnable attention sink | not in its code | not stated | None of the attention modules (Florence2Attention, Florence2SdpaAttention, Florence2FlashAttention2, or the transformer.py Attention) add… show the whole verdictNone of the attention modules (Florence2Attention, Florence2SdpaAttention, Florence2FlashAttention2, or the transformer.py Attention) add a learned per-head sink logit to the softmax denominator. |
| Gated attention output | not in its code | not stated | No sigmoid gate multiplies the attention output before the output projection in any attention class; outputs go straight from attn@v to… show the whole verdictNo sigmoid gate multiplies the attention output before the output projection in any attention class; outputs go straight from attn@v to out_proj/proj. |
| QK normalization | not in its code | not stated | The language-model attention paths (Florence2Attention family) have no q/k normalization; the only q_norm/k_norm code exists in… show evidenceThe language-model attention paths (Florence2Attention family) have no q/k normalization; the only q_norm/k_norm code exists in transformer.py's Attention class guarded by qk_norm flag, which is never set true or exposed via XVLAConfig/SoftPromptedTransformer, so it defaults off for this model. |
| Partial RoPE | not in its code | not stated | No rotary position embedding is applied at all (text model uses learned absolute positional embeddings), so there is no partial RoPE… show evidenceNo rotary position embedding is applied at all (text model uses learned absolute positional embeddings), so there is no partial RoPE mechanism. |
| YaRN RoPE scaling | not in its code | not stated | There is no rope_scaling key in the config and no RoPE-scaling code (YaRN or otherwise) anywhere in the provided source, consistent with… show the whole verdictThere is no rope_scaling key in the config and no RoPE-scaling code (YaRN or otherwise) anywhere in the provided source, consistent with the model using learned absolute positions instead of RoPE. |
| Layers without positional encoding (NoPE) | not in its code | not stated | All attention layers in the text model use the same learned absolute positional embedding scheme uniformly; there is no mixture of layers… show the whole verdictAll attention layers in the text model use the same learned absolute positional embedding scheme uniformly; there is no mixture of layers with and without positional encoding. |
| Mixture of experts | not in its code | not stated | Feed-forward blocks (fc1/fc2 in Florence2EncoderLayer/DecoderLayer, and Mlp in transformer.py) are single dense MLPs with no router or… show evidenceFeed-forward blocks (fc1/fc2 in Florence2EncoderLayer/DecoderLayer, and Mlp in transformer.py) are single dense MLPs with no router or expert selection. |
| Shared expert | not in its code | not stated | There is no MoE structure at all, so no shared-expert mechanism exists in this model. |
| Auxiliary-loss-free load balancing (selection bias) | not in its code | not stated | No router or expert-selection bias mechanism exists since the model has no mixture-of-experts layers. |
| Multi-token prediction layers | not in its code | not stated | No config key or code declares extra next-next-token prediction modules; the model predicts a single action chunk via diffusion-style… show the whole verdictNo config key or code declares extra next-next-token prediction modules; the model predicts a single action chunk via diffusion-style denoising, not multi-token text prediction. |
| SwiGLU feed-forward | not in its code | not stated | Feed-forward blocks use GELU activation (Florence2 text config activation_function is gelu, and transformer.py's Mlp uses nn.GELU), not a… show evidenceFeed-forward blocks use GELU activation (Florence2 text config activation_function is gelu, and transformer.py's Mlp uses nn.GELU), not a SiLU-gated SwiGLU structure. florence_config.text_config.activation_function = "gelu" |
filed at the root 1
“a clean, flow-matching-based VLA design relying exclusively on soft-prompted standard Transformers”
multimodal architecture 7
“By associating each hardware configuration with learnable prompt embeddings, the model can flexibly capture domain-specific knowledge”
“Our design introduces a streamlined encoding pipeline that integrates soft prompts and explicitly disentangles high- and low-dimensional…”
“we implement a cross-attention–based resampler that maps domain-specific observations into a shared representation space before feeding…”
“Each domain is assigned its own resampler and a dedicated action head, while the core Transformer backbone remains shared across all…”
“These descriptions are concatenated with the task instruction and processed by the Florence-Base encoder”
“A pretrained VLM encoder (Florence-Large (Xiao et al., 2024) in X-VLA) is used for the main vision-language stream”
“auxiliary views such as wrist-views are processed with a shared vision backbone”
6 more mentioned or evaluated
multimodal architecture “We isolate the action modality from visual–language inputs.”
multimodal architecture “domain-specific projection layers are also applied on top of multi-modal inputs to align them before being fed into the backbone”
multimodal architecture “A multimodal variant of DiT that allocates separate parameters for different input modalities and integrates them through attention”
multimodal architecture “natural language descriptions of hardware configurations are provided as additional inputs”
channel mixer “this design employs a parallel MLP-Mixer (Tolstikhin et al., 2021)–based action module alongside a pretrained VLM for vision–language…”
prediction head “This strategy addresses heterogeneity by assigning separate projection heads at the model output to map action tokens into…”
Not used 2
expert routing “we designed a heterogeneity-guided routing strategy that aimed to activate experts based on embodiment-specific cues.”
expert load balancing “To mitigate this, we give another try to introduce load-balancing regularization (Wang et al., 2024a), but the resulting rapid switching…”
filed at the root 1
“To train the velocity field, we use the OT (optimal transport) path”
learning-rate schedule 1
“use of a reduced learning rate for the soft prompts as well as for the vision–language modules”
training stability 1
“All models are adapted using our proposed two-step procedure: during the first 1,000 iterations, only the soft prompts and action heads…”
data sourcing 1
“we find it essential to adopt a DAgger-style data collection strategy (Ross et al., 2011)”
data mixture & curriculum 2
“this includes cross-domain and cross-trajectory shuffling, ensuring that the model is consistently exposed to a diverse and balanced…”
“we adopt a weighted sampling strategy combined with a carefully designed shuffling pipeline.”
1 more mentioned or evaluated
data filtering “we encourage future researchers to exercise caution when curating data for training large-scale robotics models, particularly by filtering…”
supervised fine-tuning 4
“Currently, deployment still relies on embodiment-specific adaptation, typically involving the collection of a small number of…”
“Then, we jointly optimize both the backbone and the warmed-up prompt”
“we train ACT (Zhao et al., 2023) after every 100 collected episodes, identify its failure modes, and then collect targeted demonstrations…”
“The prompt is first warmed up while keeping the pretrained weights frozen.”
2 more mentioned or evaluated
supervised fine-tuning “we adopt PEFT techniques such as Low-Rank Adaptation (LoRA)”
supervised fine-tuning “we conduct joint finetuning experiments on a mixture of downstream datasets, including Libero, BridgeData, and Calvin-ABC”
Not used 1
supervised fine-tuning “we experimented with Low-Rank Adaptation (LoRA)-style modules (Hu et al., 2022), where domain-specific adapters were introduced in…”
agentic scaffolding 1
“we provide embodiment-aware textual descriptions that encode hardware configurations and camera setups for each domain.”
filed at the root 1
“We conduct open-loop validation experiments to monitor pretraining convergence and ensure fair comparisons across different architectures…”
benchmark 1
“we carefully construct a dedicated validation set. Specifically, we sample trajectories from AGIBOT-beta (Bu et al., 2025) that are…”
other 2
filed at the root 2
“At that point, we employ swinging motions to complete the smoothing stage and then transition to stage II.”
“we decompose the folding task into two stages: (1) smoothing out the cloth from a highly disordered state, and (2) folding the smoothed…”
unfiled 15
“introducing separate sets of learnable embeddings for each distinct embodiment”
“We introduce a soft-prompt mechanism that explicitly absorbs embodiment-specific variability while preserving a shared backbone for…”
“we assign a set of learnable embeddings to each data source as Soft Prompts”
“X-VLA introduces a Soft Prompt approach with minimally added parameters”
“relying exclusively on soft-prompted standard Transformers”
“we concatenate them along with their corresponding time embeddings within the flow-matching pipeline”
“we collect demonstrations for stage I in a repetitive manner until meaningful keypoints, such as the two corners or two ends of the cloth…”
“generate a sequence of 30 anchor points that summarize the intended trajectory over the next 4 seconds”
“the absolute EEF rotation encoded using the Rotate6D representation”
“In this work, we adopt the soft prompt paradigm for robotics”
“we standardize the action space into end-effector (EEF) pose representation”
“a simple temporal downsampling strategy can help abstract action intentions and thereby facilitate more efficient pretraining.”
“we temporarily downsample demonstrations to construct abstract representations of action intentions”
1 more mentioned or evaluated
“A direct application of the Diffusion Transformer (DiT) decoder (Peebles & Xie, 2023) that generates actions conditioned on multimodal…”
Not used 1
“we initially approached this problem from the perspective of heterogeneous parameter learning. Concretely, we assigned distinct parameter…”
Sources
The curated source list for this model. "not fetched" means the URL is recorded but its text was not read in the current run.
| Source | Kind | Publisher | Status |
|---|---|---|---|
| huggingface.co/2toINF/X-VLA-Pt | model card | X-VLA authors | read |
| arxiv.org/abs/2510.10274 | technical report | X-VLA authors | read |