Deprecated: The each() function is deprecated. This message will be suppressed on further calls in /home/zhenxiangba/zhenxiangba.com/public_html/phproxy-improved-master/index.php on line 456
t-tech (T-Tech)

T-Tech

company

https://t-tech.dev

AI & ML interests

Scientific research; Natural language processing: speech analytics, search engines, dialogue systems; A family of LLMs; Speech technologies; Fraud prevention technologies; Computer vision; Recommender systems; Time series analysis

Recent Activity

elephantmipt submitted a paper about 18 hours ago

Interpreting and Steering a Text-to-Speech Language Model with Sparse Autoencoders

oltsy authored a paper 5 days ago

Learning Selective LLM Autonomy from Copilot Feedback in Enterprise Customer Support Workflows

kefirski authored a paper 10 days ago

ESSA: Evolutionary Strategies for Scalable Alignment

View all activity

Papers

Interpreting and Steering a Text-to-Speech Language Model with Sparse Autoencoders

Trust-Region Behavior Blending for On-Policy Distillation

View all Papers

t-tech 's papers 20

Submitted by

Nikita Balagansky

Interpreting and Steering a Text-to-Speech Language Model with Sparse Autoencoders

t-tech

Submitted by

Alexey Gorbatovski

Trust-Region Behavior Blending for On-Policy Distillation

t-tech

4

Submitted by

Natyren

Next Embedding Prediction Makes World Models Stronger

t-tech

Submitted by

Alexey Gorbatovski

F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare

t-tech

3

Submitted by

Olga Tsymboi

T-pro 2.0: An Efficient Russian Hybrid-Reasoning Model and Playground

t-tech

4

Small Vectors, Big Effects: A Mechanistic Study of RL-Induced Reasoning via Steering Vectors

t-tech

Submitted by

Daniil Gavrilov

Enhancing Vision-Language Model Training with Reinforcement Learning in Synthetic Worlds for Real-World Success

t-tech

Submitted by

Nikita Balagansky

Teach Old SAEs New Domain Tricks with Boosting

t-tech

1

ESSA: Evolutionary Strategies for Scalable Alignment

t-tech

Train One Sparse Autoencoder Across Multiple Sparsity Budgets to Preserve Interpretability and Accuracy

t-tech

Submitted by

Daniil Laptev

Train Sparse Autoencoders Efficiently by Utilizing Features Correlation

t-tech

2

Steering LLM Reasoning Through Bias-Only Adaptation

t-tech

Submitted by

Yaroslav Aksenov

You Do Not Fully Utilize Transformer's Representation Capacity

t-tech

Submitted by

Daniil Gavrilov

Analyze Feature Flow to Enhance Interpretation and Steering in Language Models

t-tech

2

Submitted by

Alexey Gorbatovski

The Differences Between Direct Alignment Algorithms are a Blur

t-tech

3

Submitted by

Daniil Gavrilov

Mechanistic Permutability: Match Features Across Layers

t-tech

2

Submitted by

AK

Learn Your Reference Model for Real Good Alignment

t-tech

Submitted by

AK

Linear Transformers with Learnable Kernel Functions are Better In-Context Models

t-tech

Classifiers are Better Experts for Controllable Text Generation

t-tech

PALBERT: Teaching ALBERT to Ponder

t-tech