CONVERSATIONAL AI LLMs Flashcards

1
Q

4 challenging factors in speech recognition

A
  • Variability in speech: accent, dialect, background noise
  • Domain-specific lexicon
  • Contextual understanding
  • Speech disfluency: self-corrections, repetitions
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What are important attributes of speech synthesis

A

Natural, emotional, and real-time

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Task-oriented conversations vs open domain conversations

A

Task-oriented
-clear goal
-limited knowledge
-short conversation
-predictable

open domain
-no exact goal
-unrestricted
-long-session
-unpredictable
-emotions and empathy
We have to pass the current question but also the context and rank the best response

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

How do we train LLMs

A

Train the base model and then
Use reinforcement learning from human feedback RLFH

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What are challenges of conversational AI

A
  • consistent conversations (stick with facts and persona)
  • empathic agents for user experience
  • terrible at detecting misinformation
  • using large contexts; often middle context is not helpful
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What are Hallucinations

A

When an AI generates something that is not true

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What are Intrinsic hallucinations

A

Response is not grounded in the context
obviously wrong

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What are Extrinsic hallucinations

A

Incorporates incorrect information
Harder to recognise

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

How can we improve generated responses

A

Few-shot prompting (in-context learning)
providing examples in context

Roleplay - adopting a role/function

Prompt engineering (chain of thought)
Ask model to generate the reasoning step to their answer, allows it to perform better

Finetuning
Trained on huge dataset and fine-tuned on smaller relevant data
But hard to introduce new knowledge

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Restricting context length: RAG

A

Retrieval augmented generation
Take query and only retrieve relevant context
computationally longer

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

RAG vs Fine tune

A

Start with prompt engineering

if want to improve short term memory -> RAG

if want to improve long term memory (persona or language it uses) -> Fine tune

These methods are all additive - can use all

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

How to evaluate generative LLMs

A
  • factuality of answer (faithfulness)
  • relevancy to question
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What are Foundation Models

A

Models that are capable of a range of general tasks
And then we can build things on top of it

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What do LLMs learn

A

Formal competence
- knowledge of linguistic rules and patterns;
- producing reasonable utterances

Functional competence
- understanding and using language in the world
- showing the understanding?

LLMs master formal well but functional less so

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Prompting vs fine tuning

A

Fine tuning
Using a foundational model and fine tuning it on specific data - the model changes

Prompting
Can involve only instruction or +context or +input text

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What is prompt engineering

A

the practice of developing and optimizing prompts to efficiently use LLMs for a variety of applications

17
Q

What is in-context learning

A

Prompting language model performs a task just by conditioning on input-output examples (changing prompts), without optimising any parameters
1) pertaining
2) warming up (optional eg fine tune)
3) scoring - putting together input and most feasible answer

18
Q

What is Warming up

A

Optional fine tuning for a prompting model
- Supervised in-context training: fine-tune on a a broad range of tasks
- Self-supervised in-context training: use the frozen PLM to generate some synthetic training data and fine tune on it

Model becomes better trained on ICL
But requires updating weights

19
Q

What is zero-shot learning

A

No example of expected behaviour from the model

20
Q

Prompting tips

A
  • Be specific, clear, and concise.
  • Include context.
  • Iterate: subsequent prompts are typically needed
  • Use as domain-specific instructions as possible
  • Roleplay
21
Q

Prompting problems

A
  • Can introduce bias (using context)
  • Requires domain expertise ; eg specific medical field
  • still lags behind SotA model tuning results
  • sub-optimal and sensitive
22
Q

How can we reduce Hallcuinations

A

Training it to say “I dont know”

Evidence-based response eg RAG: model has a “reality check” against real-world data
Context retrieval provides ‘evidence’: enables factual consistency

23
Q

What is self reflective RAG

A

RAG and then Validate the response to see if it is coming from the right context

24
Q

How to evaluate Retrieval LLMs

A
  • signal to noise ration of retrieved context (context precision)
  • did it retrieve all relevant information (context recall)