We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters.
Metadata: arXiv. Abstracts retain their original authors’ rights.
In the conversation
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
Likely an improvement on:
> We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters.
<https://arxiv.org/abs/2502.05171>
Amazing write up and i wish more people showed the process for discovery which is often even more interesting than the result itself
Still the result is really interesting being able to stack abstract reasoning and get better performance and the heat maps to show the prob results
The academic literature seems to be catching up:
- *[SOLAR / DUS (Kim et al., 2023)](https://arxiv.org/abs/2312.15166)* — duplicated transformer layers to build a 10.7B model that outperformed 30B parameter baselines.
- *[The Curse of Depth (2025)](https://arxiv.org/abs/2502.05795)* — explains why this works: Pre-LN causes deep transformer layers to converge toward identity functions, meaning…
Read the full comment
middle layers are where real computation happens, and duplicating them concentrates that capacity.
- *[Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach (Geiping et al., NeurIPS 2025)](https://arxiv.org/abs/2502.05171)* — takes the idea to its logical conclusion: a model trained with a single recurrent block repeated at inference time, scaling reasoning depth without adding parameters.
> nobody has tried to generalize it for example by combining the recurrence concept with next token prediction
Here you go: https://arxiv.org/abs/2502.05171
Very cool to see the effectiveness of recurrence on ARC. For those interested in recurrence, here are other works that leverage a similar approach for other types of problems:
Language modeling:
Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach https://arxiv.org/pdf/2502.05171
Puzzle solving:
A Simple Loss Function for Convergent Algorithm Synthesis using RNNs https://openreview.net/pdf?id=WaAJ883AqiY
End-to-end Algorithm Synthesis with Recurrent Networks: Logical Extrapolation Without Overthinking https://arxiv.org/abs/2202.05826
Can You Learn an Algorithm? Generalizing from Easy to Hard Problems with Recurrent Networks https://proceedings.neurips.cc/paper/2021/file/3501672ebc68a...
General:
Think Again Networks and the Delta Loss https://arxiv.org/pdf/1904.11816
Universal Transformers https://arxiv.org/abs/1807.03819
Adaptive Computation Time for Recurrent Neural Networks https://arxiv.org/pdf/1603.08983
Unfortunately I think the competitive pressures to improve model performance makes this kind of monitor-ability is short-lived. It's just not likely that textual reasoning in english ismost optimal.
Researchers are already pushing in this direction:
https://arxiv.org/abs/2502.05171
"We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach…
Read the full comment
does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters."
https://arxiv.org/abs/2412.06769
Can't monitor the chain of thought if it's no longer in a human legible format.
There is a LOT of effort in the research community currently:
1. Improving the Self-Attention in the Transformer as is, keeping the quadratic complexity, which has some theoretical advantage in principle[1]: The most hyped one probably DeepSeek's Multi-Latent Attention[15], which kind of is Attention still - but also somehow different.
2. Linear RNNs: This starts from Linear Attention[2], DeltaNet[3], RKWV[4], Retention[5], Gated Linear Attention[6], Mamba[7], Griffin[8], Based[9], xLSTM[10], TTT[11], Gated DeltaNet[12], Titans[13].
They all have an update like: C_{t} = F_{t} C_{t-1} + i_{t} k_{t} v_{t}^T with a cell state C and output h_{t} = C_{t}^T q_{t}. There's a few…
Read the full comment
tricks that made these work and now being very strong competitors to Transformers. The key here is the combination of an linear associative memory (aka Hopfield Network, aka Fast Weight Programmer, aka State Expansion...) and pushing it into a sequence with gating similar to the original LSTM (input, forget, output gate) - while here this is only dependent on the current input not the previous state for linearity. The linearity is needed to make it sequence-parallelizable, there are efforts now to add non-linearities again, but let's see. Their main benefit+downside both is that they have a fixed-size state, and therefore linear (vs Transformer-quadratic) time complexity.
For larger sizes they have become popular in hybrids with Transformer (Attention) Blocks, as there are problems with long context tasks [14]. Cool thing is they can also be distilled from pre-trained Transformers with not too much performance drop [16].
3. Along the sequence dimension most things can be categorized in these two. Attention and Linear (Associative Memory Enhanced) RNNs are heavily using Matrix Multiplications and anything else would be a waste of FLOPs on current GPUs. The essence is how to store information and how to interact with it, there might be still interesting directions as other comments show. Other important topics that go into the depth / width of the model are: Mixture of Experts, Iteration (RNNs) in Depth[17].
Disclaimer: I'm author of xLSTM and we recently released a 7B model [18] trained at NXAI, currently the fastest linear RNN at this scale and performance. Happy to answer more questions on this or the current state in this field of research.
[1] https://arxiv.org/abs/2008.02217
[2] https://arxiv.org/abs/2006.16236
[3] https://arxiv.org/pdf/2102.11174
[4] https://github.com/BlinkDL/RWKV
[5] https://arxiv.org/abs/2307.08621
[6] https://arxiv.org/pdf/2312.00752
[7] https://arxiv.org/abs/2312.06635
[8] https://arxiv.org/pdf/2402.19427
[9] https://arxiv.org/abs/2402.18668
[10] https://arxiv.org/abs/2405.04517
[11] https://arxiv.org/abs/2407.04620
[12] https://arxiv.org/abs/2412.06464
[13] https://arxiv.org/abs/2501.00663
[14] https://arxiv.org/abs/2406.07887
[15] https://arxiv.org/abs/2405.04434
[16] https://arxiv.org/abs/2410.10254
[17] http://arxiv.org/abs/2502.05171
[18] https://huggingface.co/NX-AI/xLSTM-7b
Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition.
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
As AI agents powered by Large Language Models (LLMs) become increasingly versatile and capable of addressing a broad spectrum of tasks, ensuring their security has become a critical challenge.