ReadyArt/Omega-Darkest_The-Broken-Tutu-GLM-32B
Text Generation
•
33B
•
Updated
•
26
•
•
7
Min Cho
Wayyguu
AI & ML interests
Polynomial Regression with standard deviation channels.
Recent Activity
liked
a model
about 2 months ago
ReadyArt/Omega-Darkest_The-Broken-Tutu-GLM-32B
liked
a model
2 months ago
mradermacher/Broken-Tutu-24B-i1-GGUF
liked
a model
2 months ago
lodestones/Chroma
Organizations
None yet
liked
a
model
about 2 months ago
liked
2
models
2 months ago
liked
a
dataset
3 months ago
liked
3
models
3 months ago
reacted to
Kseniase's
post with 🔥
3 months ago
Post
4251
10 new Chain-of-Thoughts (CoT) methods
CoT has long been one of the hottest techniques in AI thanks to its effectiveness and compelling core idea: encouraging models to solve complex problems through explicit intermediate reasoning steps. But usually researchers modify original CoT approach, finding tips that further improve LLMs' reasoning. That's what we're going to talk about today.
Here's a list of 10 latest enhanced CoT approaches:
1. Chain-of-Defensive-Thought -> Chain-of-Defensive-Thought: Structured Reasoning Elicits Robustness in Large Language Models against Reference Corruption (2504.20769)
Provides a few structured, defensive reasoning exemplars to improve the robustness of LLMs
2. Hybrid-CoT -> AdaR1: From Long-CoT to Hybrid-CoT via Bi-Level Adaptive Reasoning Optimization (2504.21659)
Proposes using Adaptive Hybrid Reasoning Model (AdaR1) that combines Long- and Short-CoT, and applying bi-level preference training to select effective reasoning styles
3. Semantic-level and token-level CoT -> T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT (2505.00703)
Introduces T2I-R1 text-to-image gen model, that uses semantic-level CoT for prompt planning and token-level CoT for pixel-level generation, while BiCoT-GRPO coordinates them both
4. Speculative CoT (SCoT) -> Efficient Reasoning for LLMs through Speculative Chain-of-Thought (2504.19095)
SCoT drafts multiple reasoning paths with a lightweight draft, selects the best, and uses the target model for correction - all this to reduce latency by 48–66%
5. Collaborative CoT (Co-CoT) -> Co-CoT: A Prompt-Based Framework for Collaborative Chain-of-Thought Reasoning (2504.17091)
Breaks reasoning into blocks that users can inspect, modify and re-run, promoting active engagement. An adaptation mechanism aligns outputs with diverse cognitive styles and user goals
6. XS-CoT -> Enhancing Non-Core Language Instruction-Following in Speech LLMs via Semi-Implicit Cross-Lingual CoT Reasoning (2504.20835)
It's a cross-lingual framework that integrates speech-to-text translation into reasoning, using a semi-implicit CoT approach to compress intermediate tokens. This improves non-core language responses by up to 45%
Read further in the comments 👇
If you liked this, also subscribe to the Turing Post -> https://www.turingpost.com/subscribe
CoT has long been one of the hottest techniques in AI thanks to its effectiveness and compelling core idea: encouraging models to solve complex problems through explicit intermediate reasoning steps. But usually researchers modify original CoT approach, finding tips that further improve LLMs' reasoning. That's what we're going to talk about today.
Here's a list of 10 latest enhanced CoT approaches:
1. Chain-of-Defensive-Thought -> Chain-of-Defensive-Thought: Structured Reasoning Elicits Robustness in Large Language Models against Reference Corruption (2504.20769)
Provides a few structured, defensive reasoning exemplars to improve the robustness of LLMs
2. Hybrid-CoT -> AdaR1: From Long-CoT to Hybrid-CoT via Bi-Level Adaptive Reasoning Optimization (2504.21659)
Proposes using Adaptive Hybrid Reasoning Model (AdaR1) that combines Long- and Short-CoT, and applying bi-level preference training to select effective reasoning styles
3. Semantic-level and token-level CoT -> T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT (2505.00703)
Introduces T2I-R1 text-to-image gen model, that uses semantic-level CoT for prompt planning and token-level CoT for pixel-level generation, while BiCoT-GRPO coordinates them both
4. Speculative CoT (SCoT) -> Efficient Reasoning for LLMs through Speculative Chain-of-Thought (2504.19095)
SCoT drafts multiple reasoning paths with a lightweight draft, selects the best, and uses the target model for correction - all this to reduce latency by 48–66%
5. Collaborative CoT (Co-CoT) -> Co-CoT: A Prompt-Based Framework for Collaborative Chain-of-Thought Reasoning (2504.17091)
Breaks reasoning into blocks that users can inspect, modify and re-run, promoting active engagement. An adaptation mechanism aligns outputs with diverse cognitive styles and user goals
6. XS-CoT -> Enhancing Non-Core Language Instruction-Following in Speech LLMs via Semi-Implicit Cross-Lingual CoT Reasoning (2504.20835)
It's a cross-lingual framework that integrates speech-to-text translation into reasoning, using a semi-implicit CoT approach to compress intermediate tokens. This improves non-core language responses by up to 45%
Read further in the comments 👇
If you liked this, also subscribe to the Turing Post -> https://www.turingpost.com/subscribe
- 1 reply
liked
4
models
3 months ago
mlabonne/NeuralDaredevil-8B-abliterated
Text Generation
•
8B
•
Updated
•
15.7k
•
•
220
mradermacher/Qwen3-30B-A7.5B-24-Grand-Brainstorm-GGUF
31B
•
Updated
•
2.8k
•
4
Mungert/DistilQwen2.5-DS3-0324-14B-GGUF
15B
•
Updated
•
34
•
2
chuanli11/Llama-3.2-3B-Instruct-uncensored
Text Generation
•
4B
•
Updated
•
8.43k
•
132
upvoted
a
collection
3 months ago
liked
3
models
3 months ago
reacted to
sometimesanotion's
post with ❤️
3 months ago
Post
1860
The capabilities of the new Qwen 3 models are fascinating, and I am watching that space!
My experience, however, is that context management is vastly more important with them. If you use a client with a typical session log with rolling compression, a Qwen 3 model will start to generate the same messages over and over. I don't think that detracts from them. They're optimized for a more advanced MCP environment. I honestly think the 8B is optimal for home use, given proper RAG/CAG.
In typical session chats, Lamarck and Chocolatine are still my daily drives. I worked hard to give Lamarck v0.7 a sprinkling of CoT from both DRT and Deepseek R1. While those models got surpassed on the leaderboards, in practice, I still really enjoy their output.
My projects are focusing on application and context management, because that's where the payoff in improved quality is right now. But should there be a mix of finetunes to make just the right mix of - my recipes are standing by.
My experience, however, is that context management is vastly more important with them. If you use a client with a typical session log with rolling compression, a Qwen 3 model will start to generate the same messages over and over. I don't think that detracts from them. They're optimized for a more advanced MCP environment. I honestly think the 8B is optimal for home use, given proper RAG/CAG.
In typical session chats, Lamarck and Chocolatine are still my daily drives. I worked hard to give Lamarck v0.7 a sprinkling of CoT from both DRT and Deepseek R1. While those models got surpassed on the leaderboards, in practice, I still really enjoy their output.
My projects are focusing on application and context management, because that's where the payoff in improved quality is right now. But should there be a mix of finetunes to make just the right mix of - my recipes are standing by.
liked
3
models
3 months ago
mradermacher/Uncensored_Mistral-7B_v0.3_Seaftensors-GGUF-GGUF
7B
•
Updated
•
245
•
1
DavidAU/Mistral-MOE-4X7B-Dark-MultiVerse-Uncensored-Enhanced32-24B-gguf
Text Generation
•
24B
•
Updated
•
5.29k
•
69
bartowski/uncensoredai_UncensoredLM-DeepSeek-R1-Distill-Qwen-14B-GGUF
Text Generation
•
14B
•
Updated
•
7.31k
•
27